P25 V.24/DFSI: fix startup clipping and duplicate STOP teardown

[tsawyer]

Summary

This PR hardens P25 V.24/DFSI call handling by fixing startup queue/scheduler behavior and tightening end-of-call STOP generation.

What this fixes

• Prevents clipped call onset caused by startup queue timing/space issues.
• Prevents extra STOP signaling from repeated terminator units after call state is already closed.

Changes

• Reset V.24 TX scheduler epoch at stream start (m_lastP25Tx = 0) so new calls do not inherit prior timing.
• Preserve queued TX entries on serial write failure (peek first, dequeue only on successful write).
• Avoid exact-full ringbuffer wrap ambiguity by enforcing one-byte queue headroom.
• Increase DFSI V.24 P25 TX scheduler queue sizing to absorb startup burst traffic.
• Keep normal TX queue priority in V.24 clock loop (immediate queue only when normal queue has nothing ready).
• Treat V.24 modem space as advisory during startup in host/modem write path to avoid blocking first Net->RF voice frames.
• Pass imm flag from P25 writer to modem write path.
• Only issue V.24 EOS/STOP on TDU/TDULC when a TX call is actually in progress.
• Remove temporary debug instrumentation used during diagnosis.
• Add .DS_Store ignore entry to .gitignore.

Validation

• Branch tested over-the-air with rebuilt binary from this branch.
• Call teardown now follows a single STOP path; repeated terminators no longer cause duplicate STOP behavior.
• No NAK, command = 0x... or serial write failure signatures observed in test capture window.

Notes to Devs

• Some code has already been cherry pick from my GitHub. Feel free to continue especially if code review is not desired.
• None of this code has been tested on anything other than DVMV24/Quantars.
• No CC testing has been done. We're merely scratching our own itch to get conventional working to our satisfaction.
• Finally, all code was written by Codex. I'm a php programmer not a C guy.

[gatekeep]

Because we did a cherry pick I'm closing this PR (but keeping it here for documentation purposes) in favor of #117.