Backend uses multiple Ring Containers (in form of both Buffers and Descriptor Heaps) to manage data during Command List recording. Those can severly impact performance - too small or too big Containers can cause less opportunities to parallelize CPU and GPU consumption, which impacts performance.

- Adjust used Ring Container parameters (both their size and flush threshold)
- Reduce Descriptors allocated for Texture SRVs in ResourceManager
- Consider some other adjustments (ex. using one common Ring Buffer? maybe we should let Ring Containers dynamically grow to fit all per-frame data?)