fix: surface early startup failures to clients via Initializing status

The P0 performance optimization (42s -> <1s HTTP response) introduced
silent failures: start_attempt() runs in tokio::spawn() with errors
only logged, returning is_attempt_running=true before knowing if the
attempt actually started.

This fix:
- Add `Initializing` status to ExecutionProcessStatus enum
- Create ExecutionProcess with Initializing status BEFORE spawning
- Transition to Running on success, Failed on error
- SSE automatically pushes status changes to clients

Also fixes logging:
- main.rs: Log success/failure of PATH cache warming (not just "warmed")
- shell.rs: Restore debug/warn logging for shell failures/timeouts

forge-core files changed:
- crates/db/src/models/execution_process.rs: New status + helper methods
- crates/server/src/routes/tasks.rs: Create EP early
- crates/server/src/routes/task_attempts.rs: Same pattern
- crates/services/src/services/container.rs: New start_attempt_with_process()
- crates/server/src/main.rs: Better PATH warming logs
- crates/utils/src/shell.rs: Restore shell error logging

Author: namastex888

.cargo/config.toml

@@ -32,3 +32,4 @@ rustflags = ["-C", "link-arg=-Wl,-z,max-page-size=16384"]
 # forge-core-utils = { path = "forge-core/crates/utils" }
 # ============================================================================
 
+