Per-client request rate limits over a minute, an hour and a day (closes #43)

Each client, one IPv4 address or one IPv6 /64, has its requests counted
in two buckets per window, the earlier weighted by how much of it the
window still covers, in a table of at most 20,000 clients that drops the
least recently seen. A request over SWWAF_RATE_LIMIT_PER_MINUTE, _HOUR or
_DAY (1000, 10000, 50000, or off) gets 429 before anything reaches the
app, and refused requests count. The log line gains limit_hit and the
action rate_limited. The rate limits run before the announced-size
check, so a request refused with 413 is counted too.

Deviation from SPEC.md, per the issue: the 20,000 bound and the /64 are fixed, not settings.
Judgement call: golang-lru/v2 holds the table; httprate is not used, as it reads the wall clock and does not count refused requests.
Deviation: go.mod and go.sum were written by hand from the Go checksum database, as no make target runs go mod tidy.

Model: opus-5-5

internal/proxy/client.go

@@ -48,6 +48,20 @@ func clientAddress(
 	return client
 }
 
+// ipv6GroupPrefix is the length of the IPv6 netblock that is one client.
+const ipv6GroupPrefix = 64
+
+// clientGroup is the client a request is counted toward: its IPv4
+// address, or the /64 its IPv6 address is in, since one abuser usually
+// holds a whole /64.
+func clientGroup(addr netip.Addr) netip.Prefix {
+	if addr.Is6() {
+		return netip.PrefixFrom(addr, ipv6GroupPrefix).Masked()
+	}
+
+	return netip.PrefixFrom(addr, addr.BitLen())
+}
+
 // isInside reports whether addr is in one of the netblocks.
 func isInside(addr netip.Addr, netblocks []netip.Prefix) bool {
 	return slices.ContainsFunc(netblocks, func(netblock netip.Prefix) bool {