WIP: [GPU] Logging cleanup

[echeresh]

This is related to https://jira.devtools.intel.com/browse/MFDNN-11400.

It'd be nice simplify and unify logging. This is the initial set of changes.

I plan to look into:

• Simplifying log levels, e.g. drop perf-specific levels as they are not really used AFAIK (?)
• Documenting the usage scenarios
• Moving the existing functionality to src/gpu
• Reusing spdlog (probably not much value besides prefix printing)

[rjoursler]

Simplifying log levels, e.g. drop perf-specific levels as they are not really used AFAIK (?)

I have often used this when trying to optimize IR creation time. The main issue is that the trace level introduces overhead (~30% from what I remember) and this overhead was not evenly distributed. This makes is so that performance in trace mode only loosely corresponds to performance in release mode when prioritizing optimizations. Could we add perf information (perhaps in a more compressed format) to the info level, analogous to how oneDNN verbose is mapps into spdlog levels?