Bound download expansion and escape control chars on the terminal (closes #164)

Objects fetched from the store are untrusted; several decode paths let one expand or print without limit.

- blobgen.LimitReader errors past a byte cap (not io.LimitReader silent EOF). DecodeManifest reads through caps on both compressed input and decompressed output, far above any real manifest, so json.Decode cannot buffer a compressible bomb. FetchAndDecryptBlob bounds decompression to the blob recorded uncompressed_size (not the restoring host blob_size_limit).
- downloadSnapshotDB streams straight from storage to its temp file with io.Copy, replacing two ReadAll calls that held the whole database twice.
- FetchBlob drops the per-blob Stat round-trip, its expectedSize parameter and returned size, all of which only fed a debug log.
- TTYHandler and ui.Writer escape control characters in messages, attribute keys/values, and rendered identifiers/paths before colour codes are applied, so a crafted value cannot drive the terminal.

Model: opus-4-8

internal/blobgen/limit.go

@@ -0,0 +1,49 @@
+package blobgen
+
+import (
+	"errors"
+	"io"
+)
+
+// ErrOutputTooLarge is returned by a reader from LimitReader once it has
+// been asked for more than its limit. It bounds how far an untrusted
+// compressed stream may expand, so a small, highly compressible object
+// from the store cannot decompress without limit.
+var ErrOutputTooLarge = errors.New("output exceeds size limit")
+
+// LimitReader returns a reader that yields at most limit bytes from r and
+// then fails with ErrOutputTooLarge. Unlike io.LimitReader, which reports
+// a silent io.EOF at the limit (indistinguishable from a stream that
+// simply ended), this fails, so a caller decoding or copying the stream
+// sees an error rather than a truncated value. A stream of exactly limit
+// bytes reads back cleanly to EOF; the first byte beyond it is the error.
+func LimitReader(r io.Reader, limit int64) io.Reader {
+	// remaining counts down from limit+1: the extra byte is the one that,
+	// if it ever arrives, proves the stream is longer than the limit.
+	return &limitReader{r: r, remaining: limit + 1}
+}
+
+type limitReader struct {
+	r         io.Reader
+	remaining int64
+}
+
+func (l *limitReader) Read(p []byte) (int, error) {
+	if l.remaining <= 0 {
+		return 0, ErrOutputTooLarge
+	}
+
+	if int64(len(p)) > l.remaining {
+		p = p[:l.remaining]
+	}
+
+	n, err := l.r.Read(p)
+	l.remaining -= int64(n)
+
+	if l.remaining <= 0 {
+		// The (limit+1)th byte was just read: the stream is too long.
+		return n, ErrOutputTooLarge
+	}
+
+	return n, err
+}