Fix thread safety, races, and correctness issues across drivers

[HTRamsey]

Summary

Targeted fixes for thread safety, race conditions, error handling, and correctness issues across all drivers and utilities. No new APIs, no architectural changes.

Thread safety (volatile)

• Drivers: DTR/RTS flags, mReadRequest, DEBUG, breakConfig, mStatus, mDeviceType, latency timer fields accessed across threads without visibility guarantees
• SerialInputOutputManager: DEBUG, mReadTimeout, mWriteTimeout, mThreadPriority, mReadBuffer, mWriteBuffer
• XonXoffFilter: xon field

Correctness fixes

• FtdiSerialDriver: Assign endpoints by direction + bulk type instead of hardcoded index — prevents silent read/write swap if device enumerates endpoints in unexpected order
• Ch34xSerialDriver: getStatus() validates exact return length, not just >= 0 — prevents returning stale/zeroed data on partial transfer
• Cp21xxSerialDriver: Port number validation before mIsRestrictedPort assignment
• ProlificSerialDriver: interrupt() before join() on status thread — prevents indefinite hang if thread is blocked on USB transfer
• ProbeTable: Boolean.TRUE.equals(o) instead of (boolean)o — prevents NPE if probe method returns null
• ChromeCcdSerialDriver: Use actual interface count instead of hardcoded 3
• CdcAcmSerialDriver: Guard against double-releasing same interface; null check on control endpoint count
• CommonUsbSerialPort: Null check on mReadRequest in readAsync() before queueing; getSerial() throws if port not open

Error handling

• Replace silent catch(Exception ignored){} with Log.w in all driver close paths — makes USB debugging possible
• testConnection error message includes return code (rc=)

Resource management

• UsbRequest.close() on Android O+ (was only cancelling, not closing)

Other

• static final on TAG and constant fields that were incorrectly instance-scoped
• LinkedList → ArrayList for read queue (better random-access performance)
• Fix "occured" typo in log message

Test plan

• Verify existing instrumented tests pass on connected devices
• Smoke test open/close/read/write on FTDI, Prolific, CP21xx, CH34x devices
• Verify Log.w messages appear in logcat during abnormal disconnect

[kai-morich]

looks AI generated. Any real issues you want to fix?

[HTRamsey]

Believe things like usage of RESET_PURGE_TX & RESET_HXN_TX_PIPE are backwards right? But yeah all the extra debugging stuff was AI added to help me resolve a couple issues, can remove all the debugging stuff if you want

[kai-morich]

yes, PURGE constants look reversed, which also confused me when I took over the library, but DeviceTest.purgeHwBuffers() successfully tests the functionality so I kept as is.
All else looks like cosmetics, or did I miss something important?