Check the slot's database and plugin before reusing it

[lukashes]

Slot names are global to the cluster, but slotExists matched on the name alone.
A slot of the same name in another database, or one running another output
plugin, passed the check, and START_REPLICATION then failed on every start: a
restart loop the supervisor cannot heal.

• slotExists reads database and plugin alongside current_database() and
  rejects a foreign slot with SlotMismatch, naming both sides in the log.
  Returning "no slot" instead would send the run into createSlot, which fails
  on the name clash and, on the snapshot path, leaves a marker publication behind
  on every restart.
• The drain loop no longer swallows receive errors. A broken connection there is
  the same failure the blocking receive reports, so it propagates instead of
  ending the batch as if the buffer had run dry. Nothing is lost: the batch LSN
  was never confirmed.
• client_time in the standby status update is microseconds since 2000-01-01,
  not Unix seconds, so pg_stat_replication.reply_time reads as a real
  timestamp.

Closes #97.

[github-actions]

📊 Benchmark Results

Current run is the minimum over 3 passes, compared against the base branch (main @ 03765bf), built on the same runner.

Benchmark	Baseline	Current	Δ Time	Allocs	Status
Converter DELETE	141.54μs	142.70μs	+0.8%	13 → 13	➡️
Converter INSERT	138.23μs	142.16μs	+2.8%	13 → 13	➡️
Converter UPDATE	173.13μs	172.14μs	-0.6%	22 → 22	➡️
JsonSerializer	49.90μs	50.00μs	+0.2%	3 → 3	➡️
KafkaProducer produce	232.79μs	211.54μs	-9.1%	1 → 1	🟢 faster
KafkaProducer send	273.86μs	248.01μs	-9.4%	0 → 0	🟢 faster
PgOutputDecoder	94.89μs	95.30μs	+0.4%	6 → 6	➡️
getPartitionKeyValue boolean	15.45μs	15.35μs	-0.6%	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	17.19μs	17.23μs	+0.2%	1 → 1	➡️
matchStreams found	17.23μs	17.11μs	-0.6%	1 → 1	➡️
matchStreams not found	0.06μs	0.06μs	+0.0%	0 → 0	⚪ noise

Summary: 🟢 2 faster · ➡️ 8 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.}