fix: Race Condition in Revocation Database

[edbjunwang]

This PR is to fix an potential issue that two concurrent PostgreSQL backends calling revoke() simultaneously can:

• Both pass the duplicate check (neither sees the other's entry yet)
• Both write their entries — resulting in duplicate revocation entries
• Or interleave writes, corrupting the database file

The change is to use flock() to lock the file description, when another process calls revoke() and the previous process is not done yet, the second process would return error like below:

ERROR:  Could not lock revocation database, it is already locked by another process.