fix: prevent data race between sendToWriteCh and DB close (#2249)

[adityeah8969]

Description

Fixes: #2249

Fix a data race between sendToWriteCh and db.Close(). A sender goroutine could pass the blockWrites check in sendToWriteCh while close() simultaneously closed writeCh, causing a "send on closed channel" panic.

Three changes:

• sendToWriteCh now uses a select over both writeCh <- req and HasBeenClosed() so senders bail cleanly with ErrBlockedWrites during shutdown, instead of sending on a channel that may have been closed.
• An atomic "writeChInFlight" counter tracks goroutines inside sendToWriteCh so close() knows when no sender remains.
• A new drainWriteCh() method drains and rejects any orphaned requests from the channel buffer after doWrites exits, ensuring no caller hangs on req.Wait().

Checklist

• Code compiles correctly and linting passes locally
• Tests added for new functionality, or regression tests for bug fixes added as applicable