perf(mapcache): Simplify and improve implementation of MapCache to prevent expensive reoccurring redundant map cache reads

[xezon]

This change simplifies and improves the implementation of MapCache to prevent expensive reoccurring redundant map cache reads. It also removes DEBUG_LOG in buildMapListForNumPlayers because it is a spammy log that adds stalling.

Originally, the implementation in MapCache::updateCache was a total munkee disaster. Very complicated and inefficient. This change turned out to be half refactor, half performance optimization.

Problems in MapCache were:

• Standard and User Map Cache INI loaded on every call to MapCache::updateCache, even if already loaded
• Too much different logic cramped into MapCache::loadUserMaps
• Overcomplicated logic around TheGlobalData->m_buildMapCache
• Overcomplicated and inefficient logic around std::map<AsciiString, Bool> m_seen
• Standard Maps Cache built in two code paths
• Standard Maps Cache built after existing Cache was already loaded, causing poisened Map Cache generation

All this is fixed.

Performance measurement

Measured in optimized vs2022 build, with RTS_DEBUG.

Map List Click	Original Time	Optimized Time (This Change)
System Maps	203 ms	62 ms
User Maps	313 ms	62 ms

TODO

• Replicate in Generals

[Mauller]

Looks good overall, only a few points to help clear up the logic.

[Mauller]

If we are rewriting this, should we make use of the game file system like we did with the recorder class instead of raw C file handling? It might make some of the file handling a bit cleaner.

could be someting for a followup PR at some point.

[xezon]

Maybe yes. I did not specifically touch the file handling stuff, just nearby code.

[bobtista]

I can do that as a follow up PR.

[Mauller]

Looks okay to me now, could probably be more optimised in the future but the current gains in performance are pretty decent from the current refactor.

[xezon]

Some things could be moved out of the update function into init function.

Replicated in Generals without conflicts.