fix: DPE-6485 Manage mysqld directly

[paulomach]

Issue

Mysqld_safe managed mysqld fails to restart when managed by pebble, due to mysqld_safe trapping SIGTERM, sent by pebble.

Solution

• ditch mysqld_usage, manage mysqld directly
• override configure_instance method to control restart within charm code, as mysql does not recognize pebble as a valid service manager
• refactor restart method to ditch rolling ops event defer and reuse unit recovery method (from upgrade code)

Merge after PR#607
fix #569

[shayancanonical]

less rolling ops is good!

[sinclert-canonical]

Same as in this comment. Maybe we can de-indent the whole else block 🤔

[taurus-forever]

Good to have rid of mysqld_safe. Do we remember why it was introduced the first place (to recheck regression there)?

AFAIR, in #179 the InnoDB were not able to recover in the corner cases as RESTART was not working inside mysql/shell? Is it now working?

[shayancanonical]

Good to have rid of mysqld_safe. Do we remember why it was introduced the first place (to recheck regression there)?

AFAIR, in #179 the InnoDB were not able to recover in the corner cases as RESTART was not working inside mysql/shell? Is it now working?

I believe that the RESTART still will not work inside mysql-shell. That is why I assumed that we are manually restarting during initial setup. however, it may be a problem in upgrades if we expect mysql-shell to restart mysqld but cannot