[VibeSWE]

Description

Counts requests previously accepted arbitrarily large numBins, and ingestion checked the field limit before adding the request's columns. A request could therefore allocate enormous bin vectors or grow a stream past its configured schema limit.

• Restrict explicit numBins to 1–10,000 on filtered and unfiltered counts requests. Check automatically chosen bins before allocation or manifest reads, and absorb remainders into the final bin instead of allocating an extra bin.
• Enforce P_DATASET_FIELD_COUNT_LIMIT against distinct existing and incoming fields under the stream metadata lock. Include generated fields, header fields, and typed siblings; cover every user event processed through the shared HTTP, OTel, and Kafka path.
• Check the union of partitioned request schemas before processing events. Return the existing HTTP 400 field-limit error with the actual count and configured limit.

Compatibility

The existing field-limit setting and default of 1,000 are unchanged. Existing over-limit streams remain loadable and readable, but ingestion is rejected even when it adds no fields. Schema recovery and query refresh stay unrestricted; internal streams retain their existing behavior.

New header fields on otherwise known-schema events are now committed for dynamic streams. Filtered counts keep their existing SQL interval behavior. Concurrent partitioned requests retain existing partial-ingestion semantics; commit-time checks still prevent the local schema from exceeding its limit.

Validation

Backend-only validation with four build jobs, after rebasing onto current main:

• cargo test --locked --lib -- --test-threads=4: 1,240 passed, 3 ignored, no failures.
• cargo clippy --locked --all-targets: passed with existing repository warnings.
• Rustfmt checks on all five changed files and git diff --check.

Added 24 regression tests covering bin boundaries and extreme values, HTTP 400 responses, merged field counts, generated/header/typed-sibling fields, request unions, existing wide schemas, and concurrent schema writers.

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}