perf(spanner): eliminate per-RPC regex operations in metrics layer

[olavloite]

Replaces regular expression matching across hot RPC metrics pathways with direct string searching and delimiter-based parsing:

• Extracts GFE and AFE Server-Timing latencies using delimiter checks and ASCII digit accumulation, avoiding per-RPC regex instantiation and string allocations.
• Replaces named-capture regular expressions in resource name parsing with positional path segment inspection.
• Replaces regex parsing in request ID operation extraction with a single-pass delimiter scan.
• Updates latency checks in the interceptor and tracer to explicitly check for numeric types, ensuring 0 ms latencies are recorded accurately instead of misclassified as connectivity errors.

[gemini-code-assist]

Code Review

This pull request refactors metrics tracking and tracing in the Spanner client to eliminate regex allocations and improve performance. Specifically, it replaces regex-based parsing with manual string splitting and character-by-character parsing for extracting instance attributes, operation requests, and server-timing latencies (GFE and AFE). It also updates latency checks to explicitly verify number types, enabling support for zero-latency values, and adds comprehensive unit tests to validate these optimizations and edge cases. There are no review comments, so I have no feedback to provide.