fix(rpc): reduce default limits to prevent resource exhaustion

[rswanson]

Summary

• Reduce max_blocks_per_filter default from 10,000 to 2,000
• Reduce max_tracing_requests default from 25 to 10

Context

Under moderate load, wide eth_getLogs queries (up to 10,000 blocks) monopolize database connections for extended periods, especially with the streaming path that issues one query per block. Combined with the cold storage write-drain behavior (all reads must complete before each write), a few large log queries can stall the entire RPC server.

Similarly, tracing/debug requests are CPU and memory intensive. 25 concurrent tracing requests can saturate the blocking thread pool and compete with state queries for resources.

Both limits remain overridable via SIGNET_RPC_MAX_BLOCKS_PER_FILTER and SIGNET_RPC_MAX_TRACING_REQUESTS environment variables for deployments that need the higher limits.

Test plan

• cargo +nightly fmt -- --check
• cargo clippy -p signet-rpc --all-targets
• cargo t -p signet-rpc (35 tests + 6 doctests pass)

🤖 Generated with Claude Code

[rswanson]

Code review

No issues found. Checked for bugs and CLAUDE.md compliance.

🤖 Generated with Claude Code

_{- If this code review was useful, please react with 👍. Otherwise, react with 👎.}

[Evalir]

ok with testing this for now, but we should adjust this later

[prestwich]

the streaming path that issues one query per block

which path is this?

[rswanson]

holding off on this pending some other tests first, if rpc perf is still bad we can explore this. setting to draft

[rswanson]

going to close this for now, other fixes have been sufficient.