[Feature Request] Redesign token caching in MSAL.NET

[bgavrilMS]

P1 Requirements

• Memory limits on internal cache, ideally LRU based [Feature Request] Telemetry - cache size for in-memory cache #4197
• No cache serialization in memory cache
• Analyze the data for cache utilization and consider switching to shared cache + max memory size by default
• Avoid cache race conditions. It is acceptable that new cache API is different from the old, as long as the actual data remains backwards compatible
• Investigate using R9 memory cache in MSAL https://domoreexp.visualstudio.com/R9/_git/SDK/pullrequest/800604
• When similar token requests are made (e.g. N threads asking for the exact same token), do not make N requests to the STS. Instead, block.

P2 requirements

• Allow configuration of distributed caches from a lightweight package
• Support for multi-level caches [Feature Request] [SPIKE] Add support for multi-layered caching #2845