Standardize sfs log levels

[irq0]

Let's standardize sfs log levels. This came up during the devs call 2023-10-25.

Describe the solution you'd like

Define our own log subsystem to be able to adjust logging independent of the rest of RGW.

Define constants to group ceph log levels to something more human understandable.

From the devs meeting we roughly like to have:

• ERROR (-1)
  • Something unrecoverable. Must always go to log irregardless of log level set by user.
  • Database corruption, ceph_abort
• Startup Message (0 or 1)
• Important, Warning (0)
• Info (1)
• Verbose (10)
  • GC ran
  • we did a thing and found it weird enough to log
• Debug (15)
  • entered a function, left function
  • we wrote stuff to disk
• Trace (> 20)

Describe alternatives you've considered

Additional context

Running benchmarks with --debug-rgw 10 instead of --debug-rgw 0 takes time: 11h -> 17h