feat: ModeStream (Redis-like append-only log)

[Code0987]

Summary

New keyspace mode ModeStream: a Redis-like append-only log with owner-minted millis-seq ids and one opaque []byte payload.

Design: docs/design/2026-09-08-mode-stream.md

What landed vs the design

• Engine / client / Cache RPCs: XAdd, XRange, XRevRange, XLen, XDel, XTrim; existing Delete(name) tombstones the stream
• Exclusive start is a leading ( on start; end is always inclusive
• Owner mints ids; non-owner XAdd uses peer StreamAdd and returns the id
• Snapshot fan-out (FlagStream + Value[0] S); inbox A/D/T stays owner-only
• Entry.Flags widened to uint64; FlagStream = 1<<32
• XRevRange uses the same start/end window as XRange (newest first) — not Redis high-then-low argument order
• Hard cap 4096 → stream full; client rejects int32 overflow on XTrim / range count
• sc xadd|xlen|xrange|xrevrange|xdel|xtrim
• Not in this PR (design non-goals): consumer groups, blocking XREAD, caller-chosen ids, Lab widget, examples/stream

Product docs

Updated in this PR: docs/API.md, api/openapi/cache.openapi.yaml + docs/api snapshot (0.14.0), PLAN.md, docs/OPERATIONS.md, docs/CLUSTER_FLOWS.md, README.md, cmd/sc help/README, docs/design/README.md.

Test plan

• go test ./... (local, green)
• Local micros: StoreGetHit 2 allocs/op, EngineGetHit 15 allocs/op (flat)
• CI test + bench; read <!-- supercache-bench-comment -->

Bench risk

• Hot path? Get-hit / StoreGetHit: no if Stream is not on those benches.
• Flags uint32→uint64 widens store.Entry by 4 bytes. Could move StoreGetHit ns/op; must not raise Get-hit / StoreGetHit allocs/op.
• No new scbench cells.
• Gate: shared smoke ±10%; Get-hit allocs flat.

[github-actions]

SuperCache bench vs main

Both sides ran once on this same GitHub runner (main, then this PR). Not a merge gate.

Baseline: dab9d3c (2026-09-08T08:30:11Z) → this run: dab9d3c (2026-09-08T08:29:24Z)

scbench smoke

cell	ops/s (main → PR)	Δ ops/s	p50	p99
get/hit n=1 c=1 uniform	8574 → 8742	+2.0%	116µs → 114µs	195µs → 189µs
get/hit n=1 c=10 uniform	33495 → 33634	+0.4%	271µs → 268µs	823µs → 821µs
get/hit n=3 c=10 uniform	38766 → 39093	+0.8%	227µs → 225µs	892µs → 897µs
miss/miss-cacheonly n=1 c=1 uniform	8540 → 8424	-1.4%	116µs → 117µs	213µs → 218µs
miss/miss-cacheonly n=1 c=10 uniform	35521 → 33617	-5.4%	254µs → 268µs	804µs → 881µs
set/put n=1 c=1 uniform	8361 → 8484	+1.5%	117µs → 116µs	223µs → 208µs
set/put n=1 c=10 uniform	32543 → 33229	+2.1%	279µs → 274µs	851µs → 815µs

Δ ops/s: green-ish is higher throughput. Latency: lower is better.

go test -bench (ns/op)

benchmark	ns/op (main → PR)	Δ	allocs/op
BenchmarkEngineDelete	783 → 847	+8.1%	6 → 6
BenchmarkEngineGetHit	1234 → 1252	+1.5%	15 → 15
BenchmarkEngineGetHitParallel	1312 → 1199	-8.6%	15 → 15
BenchmarkEngineGetMissCacheOnly	860 → 897	+4.3%	14 → 14
BenchmarkEngineGetMissCacheOnlyParallel	635 → 635	+0.0%	14 → 14
BenchmarkEngineGetMissLoadThrough	1609 → 1620	+0.7%	24 → 24
BenchmarkEngineGetMissLoadThroughParallel	1873 → 1823	-2.7%	23 → 23
BenchmarkEnginePut	1582 → 1623	+2.6%	18 → 18
BenchmarkEnginePutParallel	1386 → 1458	+5.2%	18 → 18
BenchmarkStoreBitSet	113 → 113	+0.2%	1 → 1
BenchmarkStoreCGetHit	73 → 77	+5.5%	0 → 0
BenchmarkStoreCIncr	101 → 105	+3.3%	1 → 1
BenchmarkStoreCMSIncr	121 → 123	+1.3%	0 → 0
BenchmarkStoreCMSIncrN	121 → 121	-0.1%	0 → 0
BenchmarkStoreGeoAdd	9532 → 10127	+6.2%	2 → 2
BenchmarkStoreGeoPosHit	89 → 89	+0.2%	0 → 0
BenchmarkStoreGeoPosMiss	87 → 87	+0.3%	0 → 0
BenchmarkStoreGetHit	292 → 293	+0.5%	2 → 2
BenchmarkStoreGetHitParallel	357 → 346	-3.0%	2 → 2
BenchmarkStoreHGetHit	91 → 92	+0.9%	1 → 1
BenchmarkStoreHGetMiss	85 → 84	-0.5%	0 → 0
BenchmarkStoreHLLAdd	100 → 103	+3.2%	0 → 0
BenchmarkStoreHSet	10570 → 10980	+3.9%	3 → 3
BenchmarkStoreJsonSet	26797 → 26757	-0.1%	207 → 207
BenchmarkStorePut	304 → 343	+12.7%	2 → 2
BenchmarkStorePutParallel	352 → 362	+2.9%	2 → 2
BenchmarkStoreSetAdd	9213 → 9169	-0.5%	2 → 2
BenchmarkStoreSetContainsHit	87 → 88	+0.3%	0 → 0
BenchmarkStoreSetContainsMiss	84 → 84	-0.5%	0 → 0
BenchmarkStoreTopKAdd	996 → 992	-0.4%	20 → 20
BenchmarkStoreVAdd	107749 → 108433	+0.6%	512 → 512
BenchmarkStoreVSim	89438 → 90059	+0.7%	530 → 530
BenchmarkStoreZAdd	9862 → 9552	-3.1%	2 → 2
BenchmarkStoreZScoreHit	86 → 86	-0.4%	0 → 0
BenchmarkStoreZScoreMiss	86 → 86	+0.2%	0 → 0

_{Posted by CI bench job. Not a merge gate.}