refactor: rework the worker manager for performance

[ToddyTheNoobDud]

Changes

Added this.workerToGuilds = new Map() on constructor, this goes from an O(n) to a O(1), providing better speed

Added this.workersById = new Map(), this also enables O(1) lookups for more speed

Reworked scaleWorkers(), this removes the excessive interactions of filtering and array creations, converted to a for const with a single pass interaction

Added caching on scalingConfig to avoid repeating objects acess

Added batching to stats updates (every 100ms) to reduce overhead

getBestWorker has been rewrited to use the least-loaded worker instead of using round-robin for selecting workers.

Improved the failure history to use a slice(-5) instead of shift(), slice uses O(k), beign slighly more performant than shift()

Why

This should reduce the garbage collection overload/pressure, also should improve the workers distribution for loading and reduce the overhead while maintaining more speed

Checkmarks

• [ X ] The modified endpoints have been tested.
• [ X ] Used the same indentation as the rest of the project.
• [ X ] Still compatible with LavaLink clients.

Additional information

i request for an actual testing on this before "merging" if this happens, i want to check if this actually helps reducing the cpu and memory load, any feedback will be good