Fix SIGHUP reload coordination race condition

[peteski22]

Summary

• Add atomic flag coordination between SIGHUP signal handler and main reload loop
• Prevent race condition where subsequent SIGHUP signals would be ignored after first reload
• Add comprehensive test suite with stress testing for concurrent signal handling

Changes

• Add reloadState type with atomic flag to coordinate reload operations
• Update handleSignals to use atomic CompareAndSwap for thread safety
• Ensure reload flag is properly reset after completion
• Add new test cases covering edge cases and concurrent scenarios

Testing

Race condition testing shows no issues with concurrent access:

go test -race -count=100 ./cmd -run "TestDaemon_DaemonCmd_HandleSignals"                         
ok      github.com/mozilla-ai/mcpd/v2/cmd    76.651s

go test -race -count=100 ./cmd -run "TestDaemon_ReloadFlagManagement"                          
ok      github.com/mozilla-ai/mcpd/v2/cmd    54.903s