ProcedureCallError: preserve distinction between guest and host errors

[mahmudsudo]

Description of Changes

Addresses #5660 (follow-up from #5640).

Previously, ProcedureCallError combined all execution failures into InternalError(String), lumping guest module panics/traps together with host internal errors. Consequently, procedure error logging in the HTTP API handler was temporarily downgraded to log::info! to avoid false positive host alerts triggered by guest panics.

This PR:

1. Adds a GuestPanic(String) variant to ProcedureCallError in crates/core/src/host/module_host.rs.
2. Updates InstanceCommon::call_procedure in crates/core/src/host/wasm_common/module_host_actor.rs to return ProcedureCallError::GuestPanic when guest execution encounters a panic/trap, while preserving ProcedureCallError::InternalError for host-side failures (such as return value deserialization failure).
3. Updates map_procedure_error in crates/client-api/src/routes/database.rs to:
   • Log ProcedureCallError::GuestPanic at log::debug! level (preventing false alarm host alerts).
   • Log ProcedureCallError::InternalError at log::error! level (ensuring host alerts fire on internal host failures).
4. Adds unit test test_map_procedure_error_distinguishes_guest_panic_and_internal_error.

API and ABI breaking changes

No breaking changes to the external client wire protocol or HTTP status codes. Internal Rust enum ProcedureCallError in spacetimedb_core::host received an additional GuestPanic variant.

Expected complexity level and risk

1 - Low risk, targeted error variant separation and log level correction.

Testing

• Added unit test test_map_procedure_error_distinguishes_guest_panic_and_internal_error in crates/client-api/src/routes/database.rs
• Verified ProcedureCallError variant handling and log levels

[CLAassistant]

All committers have signed the CLA.

[mahmudsudo]

@JasonAtClockwork