Capture the slot's consistent point and split connect from streaming

[lukashes]

Groundwork for the initial snapshot (#49). A consistent snapshot needs the slot's exported consistent point and a seam to run the snapshot between slot creation and START_REPLICATION. This PR adds both.

• replication_protocol.zig: read the CREATE_REPLICATION_SLOT result row and keep consistent_point, the slot's start LSN. Captured only when we create the slot; an existing slot leaves it null.
• source.zig: split connect into connectAndEnsureSlot and beginReplication so a caller can act between the two. connect stays as a thin wrapper for callers that pass an explicit start LSN (tests).
• main.zig: stream from the captured consistent point, falling back to 0/0 when the slot already existed.
• Integration test: a freshly created slot exposes a valid consistent point and streams from it.

No behavior change on its own: for a fresh slot, 0/0 already resolves to the same position. The value is the captured point and the connect split, which the snapshot PRs build on.

[github-actions]

📊 Benchmark Results

Current run is the minimum over 3 passes, compared against the base branch (main @ b041d9c), built on the same runner.

Benchmark	Baseline	Current	Δ Time	Allocs	Status
Converter DELETE	145.47μs	127.69μs	-12.2%	13 → 14	🟢 faster
Converter INSERT	145.72μs	127.62μs	-12.4%	13 → 14	🟢 faster
Converter UPDATE	175.01μs	174.93μs	-0.0%	22 → 23	➡️
JsonSerializer	49.92μs	50.62μs	+1.4%	3 → 3	➡️
KafkaProducer produce	243.26μs	211.04μs	-13.2%	1 → 1	🟢 faster
KafkaProducer send	253.71μs	256.10μs	+0.9%	0 → 0	➡️
PgOutputDecoder	97.04μs	97.68μs	+0.7%	6 → 6	➡️
getPartitionKeyValue boolean	15.76μs	15.81μs	+0.3%	1 → 1	➡️
getPartitionKeyValue integer	0.04μs	0.04μs	+0.0%	0 → 0	⚪ noise
getPartitionKeyValue not found	0.03μs	0.03μs	+0.0%	0 → 0	⚪ noise
getPartitionKeyValue string	17.59μs	17.54μs	-0.2%	1 → 1	➡️
matchStreams found	17.54μs	17.71μs	+1.0%	1 → 1	➡️
matchStreams not found	0.05μs	0.06μs	+9.2%	0 → 0	⚪ noise

Summary: 🟢 3 faster · ➡️ 7 neutral · ⚪ 3 ignored (sub-μs)

_{Thresholds: <1μs ignore · 1–20μs 15% · 20–50μs 10% · ≥50μs 5%. Measured on a shared CI runner — treat small deltas as noise. Informational only; this check never fails the build.}

[lukashes]

/bench