fix: avoid trace executor panics

[umeeSthein]

Summary

• Replace production unwrap() calls in the trace executor with error handling.
• Return default tracer serialization failures through trace/RPC errors instead of panicking.
• Preserve partial block trace results when mux inspector recreation fails.

Tests

• cargo fmt --all --check
• cargo check -p debug-trace-server --message-format=short
• cargo test -p debug-trace-server tracing_executor --message-format=short

[flyq]

Closing: this no longer applies to main, and the three unwrap()s it targets cannot fail there. serde_json::to_value on DefaultFrame is infallible (derived Serialize; storage keys serialize as strings), and the per-transaction MuxInspector::try_from_config rebuilds from the same config that already built successfully before the loop (bin/debug-trace-server/src/tracing_executor.rs:209). The trace executor's error model has also changed since: a failure aborts the trace with a typed TraceError instead of returning partial TraceResult::Error entries (tracing_executor.rs:182), which the partial-result change here would go against. Thanks for the contribution.