perf(table): reuse partition path formatting plan

[fallintoplace]

Why

PartitionToPath runs for every partition path written by the fanout and clustered writers. Each call rebuilt the same schema-dependent formatting state:

• resolved partition field types
• transform result types
• escaped field names
• output buffer estimate

Only the partition values change between calls.

What changed

• Reuse the formatting state from the existing partition extraction plan.
• Use the plan in fanout and clustered writers.
• Keep value formatting, URL escaping, and public PartitionToPath behavior unchanged.
• Add parity coverage for escaped names, timestamps, dropped source fields, and unknown transforms.

There are no public API changes.

Benchmark

Apple M1 Pro, 64 partition fields:

Path	Time	Memory	Allocs
PartitionToPath	7.3 us/op	9.62 KB/op	67/op
Reused plan	2.8 us/op	3.10 KB/op	66/op

Tests

• go test -count=1 -skip '^TestCreateAzureBucketManagedIdentityCredentialCalled$' ./...
• go test -race -count=1 . ./table
• go run github.com/golangci/golangci-lint/v2/cmd/golangci-lint@v2.8.0 run --timeout=10m