Refactor IO "read" core logic

[mgravell]

This is a partial implementation of the IO core rewrite, that just migrates the reader logic, without being the entire "big bang"; this is intended to stabilize multiple issues associated with the existing code.

• Core read
  • Bring RESPite lib into project with RespReader core (pre-existing, well-tested work)
  • Remove pipe
  • Replace primary “read” loop with read-loop from RESPite
  • Implement based write buffer to shim write-loop
  • Implement (temporary) shim from inbound RespReader to outgoing RawResult
  • Create unit-test infrastructure for result processors
  • Whittle through result-processors until all RawResult code removed, with unit tests for each
    • ideal AI candidate
  • Remove RawResult entirely (including result processor shim)
  • Remove Pipelines.Sockets.Unofficial if possible

Later planned stages (separate release schedule):

• Core write
  • Migrate RespWriter core into RESPite
  • API to enable messages to be pre-written (recycling) at construction, and implement - noting that EVAL[SHA] and MULTI/EXEC may need consideration (mostly MULTI/EXEC; EVAL[SHA] can probably be simplified greatly, if we can re-introduce the N in “this buffer is N messages”)
  • Migrate from “message” as the primary payload to “message writer of T” with “state T”
• Advanced write
  • Migrate existing message creation code to use the Roslyn generator bits with generated message writers
  • Migrate batch logic as per RESPite
• Extensibility API
  • Expose parser/writer API for external consumers
  • migrate NRedisStack to use parser/writer API