fix(chain_plugin): close and flush database on stop-replay-at-block exit (fixes #3710)

[substeem]

Description

When running steemd with --replay-blockchain --stop-replay-at-block=<N>, the node previously exited via exit(EXIT_SUCCESS) without closing the chain database. Because the database was not closed, the shared memory file was not flushed to disk cleanly, causing the replayed state up to block <N> to be lost on the next startup.

Changes

• Explicitly call my->db.close() before calling exit(EXIT_SUCCESS) in chain_plugin::plugin_startup().
• Add log messages confirming database closure and shared memory flush.

Closes #3710

[only-dev-time]

This is not fully tested!
I also used this for db comparison (commit 47f75cd). I noticed that an in-memory replay can lead to duplicate entries in certain situations if the database is closed before the programme exits.

Example:

comment_index:
--- Size: DB1: 4181982    | DB2: 4181982   
--- Progress finished (2090991/4181982)

#3710