Route GORM's logger through slog and bound it (closes #178)

GORM's default logger printed the fully interpolated SQL to standard
output on every statement that returned an error, including a plain
record-not-found. On /webhook/{uuid} and on the login form the
interpolated parameter is client-chosen and unbounded, so an
unauthenticated client sized the operator's log, one line per request,
at no level the operator could turn down.

Every gorm.Open in the service now installs internal/gormlog, a
gormlogger.Interface over the service's *slog.Logger. Its lines take
the level the operator set and the handler internal/logger selected; a
record-not-found is not logged as an error, since it is the expected
outcome on both of those paths and each handler already records its
own miss at DEBUG without the SQL; slow statements are kept at WARN
above the same 200ms threshold GORM used; and every value it emits is
spent through an encoded-byte budget.

Trace orders its cases exactly as GORM's own Trace orders them --
error-that-is-not-a-miss, then slow, then routine -- so a statement
that both missed and ran slow is still reported as slow. Ordering the
drop first would have made this adapter strictly less observant than
the IgnoreRecordNotFoundError option it was chosen over, on the two
lookups the issue is about, and a miss is the statement most likely to
be slow.

That budget is internal/middleware's truncateLogField, moved to a new
internal/logfield package now that a second writer needs it. The move
is unchanged logic. MaxAccessLogLineBytes bounds a GORM line too, and
internal/gormlog asserts each line against the constant directly.

The third gorm.Open, in the archive writer, was not named in the issue
and had the same default. All three sites are pinned independently:
internal/handlers covers the main and per-webhook databases,
internal/delivery covers the archive writer, whose type is unexported.
Reverting any one of the three to a bare &gorm.Config{} fails the
suite.

The flood test's per-line and volume assertions were vacuous, because
the replaced default logger wrote only to a buffer while everything
else went to the captured stdout. It now tees to stdout as GORM's real
default does, so a reverted call site lands in the same capture and
those assertions measure the whole writer set.

README: the ceiling now covers GORM, and the writers it does not cover
are re-derived by measuring rather than by reading. fx's console
logger and the Go runtime write to standard error. net/http's nil
ErrorLog is not a separate writer at all -- slog.SetDefault redirects
the log package's default logger into internal/logger's handler, so
those lines arrive on standard output at INFO. A handler panic reaches
that same path because chi's Recoverer crashes before writing, which
is filed as #187 and is also the widest line the service can write, at
2,772 bytes against the stated 2,560.

internal/logfield/logfield.go

@@ -0,0 +1,140 @@
+// Package logfield bounds a client-supplied value against what the log
+// handler will actually emit for it, so a line's size is set by this
+// service rather than by the client that provoked it.
+//
+// It lives outside internal/middleware because more than one writer
+// needs it: the access log, and the GORM adapter in internal/gormlog,
+// which logs SQL with the client-chosen parameters interpolated into
+// it. One budget, one implementation.
+package logfield
+
+import (
+	"strings"
+	"unicode"
+	"unicode/utf8"
+)
+
+const (
+	// MaxBytes is the default budget for a field whose value the
+	// client supplies outright. It is spent in ENCODED bytes (see
+	// Truncate), so 512 still holds a real browser's User-Agent whole
+	// — those are plain ASCII, which encodes one byte for one — while
+	// a value built from characters the encoder escapes keeps a
+	// shorter prefix. That is the intended trade: 500 quotation marks
+	// are not a debugging asset.
+	MaxBytes = 512
+
+	// truncationMarker is appended to any value Truncate cut, so a
+	// short value and a truncated one cannot be confused. It is
+	// charged on top of the budget, not inside it.
+	truncationMarker = "[truncated]"
+)
+
+// EncodedBytes is what r costs on the line once the log handler has
+// escaped it, taking the worse of the two handlers internal/logger
+// configures.
+//
+// slog's JSON handler escapes quote, backslash, newline, carriage
+// return and tab to two bytes each, and every other C0 control plus
+// LINE SEPARATOR and PARAGRAPH SEPARATOR to a six-byte \u escape; it
+// passes every other rune through as its own UTF-8. Its text handler
+// quotes with strconv.Quote, which spells a non-printable rune below
+// U+10000 as \uXXXX but one at or above U+10000 as \UXXXXXXXX — ten
+// bytes, not six. The text handler is therefore the worse of the two
+// for every non-printable rune, and by four bytes apiece for the
+// 955,086 unassigned, private-use and format code points on planes 1
+// to 16.
+//
+// Charging ten there is what makes the stated per-line ceiling hold
+// for the tty handler as well: U+1000C encodes as F0 90 80 8C, every
+// byte >= 0x80, which httpguts.ValidHeaderFieldValue accepts and
+// net/textproto does not strip, so a header can be filled with them.
+//
+// Both handlers pass printable runes through as their own UTF-8, so
+// unicode.IsPrint separates the escaped cases from the plain ones for
+// either handler.
+func EncodedBytes(r rune) int {
+	const (
+		// A backslash and the character itself.
+		shortEscapeBytes = 2
+		// \uXXXX, which is also the width of \u00XX.
+		escapedRuneBytes = 6
+		// \UXXXXXXXX, strconv.Quote's spelling of a non-printable
+		// rune outside the basic multilingual plane.
+		escapedAstralRuneBytes = 10
+		// The first code point strconv.Quote spells with \U.
+		firstAstralRune = 0x10000
+	)
+
+	switch {
+	case r == '"' || r == '\\' || r == '\n' || r == '\r' || r == '\t':
+		return shortEscapeBytes
+	case !unicode.IsPrint(r) && r >= firstAstralRune:
+		return escapedAstralRuneBytes
+	case !unicode.IsPrint(r):
+		return escapedRuneBytes
+	default:
+		return utf8.RuneLen(r)
+	}
+}
+
+// Truncate caps s at maxBytes of ENCODED output, marking the value
+// when it cuts.
+//
+// Budgeting raw bytes would not bound the line. Escaping only ever
+// grows a value, so a raw budget spent on characters the encoder
+// escapes buys a field several times its nominal size — and the line
+// is the thing an operator is told to multiply by their request rate.
+// Charging each rune what it will actually cost is what makes the
+// stated ceiling true rather than merely larger. The visible
+// consequence is that an escape-heavy value keeps a shorter prefix
+// than a plain one, which is the correct trade.
+//
+// The result is always valid UTF-8. A cut on a byte boundary can split
+// a multi-byte rune, and a header — or a SQL literal — can carry bytes
+// that were never valid UTF-8 to begin with; both are dropped rather
+// than kept, since an encoder would otherwise spend six bytes
+// replacing each one.
+func Truncate(s string, maxBytes int) string {
+	// No rune encodes to fewer bytes than it occupies, so nothing past
+	// maxBytes raw can fit the budget. Slicing first bounds the scan
+	// below to the budget rather than to the size of the value the
+	// client sent.
+	window, cut := s, false
+	if len(window) > maxBytes {
+		window, cut = window[:maxBytes], true
+	}
+
+	var (
+		kept  strings.Builder
+		spent int
+	)
+
+	for i := 0; i < len(window); {
+		r, size := utf8.DecodeRuneInString(window[i:])
+		if r == utf8.RuneError && size == 1 {
+			i += size
+
+			continue
+		}
+
+		cost := EncodedBytes(r)
+		if spent+cost > maxBytes {
+			cut = true
+
+			break
+		}
+
+		spent += cost
+
+		kept.WriteString(window[i : i+size])
+
+		i += size
+	}
+
+	if !cut {
+		return kept.String()
+	}
+
+	return kept.String() + truncationMarker
+}