Better caching of class attributes that are unlikely to change.

[th3w1zard1]

Purpose of this PR is to cache common lookups when they're unlikely to have been changed.

• class ResourceIdentifier: This now is a frozen data class instead of inheriting from NamedTuple. This allows us to define the filename str representation on init and look it up when needed (used frequently in __hash__ and __str__. reminder that hashing any object happens in __eq__ calls and in collection objects like lists/dicts.
• utility.system.path - str() representations cached. This used to be especially problematic for CaseAwarePath instances in pykotor.tools.path as any case-insensitive path instance would have get_case_sensitive_path on every call.

probably a few other places that could use @lru_cache for recursion-specific logic, much can be done still.

[th3w1zard1]

it seems most of this PR was unintentionally merged already during the fiasco of what was once bleeding-edge. I'll go ahead and merge this to complete the process.