…-centric

- introduce `Resp2Type` and `Resp3Type` shims (`Resp2Type` has reduced types)
- mark existing `Type` as `[Obsolete]`, and proxy to `Resp2Type` for compat
- deal with null handling differences
- deal with `Boolean`, which works very differently (`t`/`f` instead of `1`/`0`)

remove RedisResult.Type from Shipped.txt

- handle RESP3 types
- handle [+|-]{inf|nan}
- avoid alloc when parsing doubles

made the RedisResult constructor non-public (fix accidental API)

remove RawResult.Type; fix broken loop

format

incorrect attribute check

- nomenclature: MultiBulk => Array
- efficiency: use bit-packing to RESP2 type conversion is a bit mask

fix RawResult.HasValue

fix null array return (EmptyMultiBulk is no longer helpful)

add a ToString to Replica (other bits were local test setup issues)

simplify switch in TryParseDouble

- protocol configuration parsing
- avoid inbuilt equality/comparison operations on Version
- rules for when to try resp3

configuration documentation

words

fix ConfigurationOptions.Clone

actually connect via RESP3

demand redis6 in the RESP3 connect test

remove unnecessary directives

Lua results

more tests and tweaks to fix tests

simplify handshake

fallback connect

move SETNAME back into HELLO message; add lots of documentation about *why*

DEBUG PROTOCOL tests; some failures to look at

fix protocol tests

add missing "hide me" attribs

add docs and release notes

tyop

redundant

re-enable to get server-maintenance notifications

- ConnectWithBrokenHello is inconclusive if not a v6 server
- allow non-RESP3 tests on non-v6 servers

"DEBUG PROTOCOL" tests are inconclusive on non-v6

fix TryConnect (CLIENT ID) not always available

save all

counting is hard

expose IAsyncEnumerable on ChannelMessageQueue (#2402)

* expose IAsyncEnumerable on ChannelMessageQueue
fix #2400

* PR number

* move ChannelMessageQueue.GetAsyncEnumerator to shipped

LUA conversions

version Lua RESP conversions

true/false handling depends on setresp(3)

revert "if" split in ResultProcessor

use enum for RedisProtocol

reinstate parameterless RedisResult .ctor

fix resp 2/3 inversion snafu from enumification

fix resp dependent connection reuse issue

add failing Execute test re RESP2 vs RESP3 delta

ValuePairInterleavedProcessorBase should auto-handle responses that have become jagged in RESP3

pattern match is easier to read here

- move IsResp3; that is a PhysicalConnection thing, not a ServerEndPoint thing
- allow RawResult to know whether it is RESP3; involved moving some flags (which removes a bit hack we were using, so: yay)
- make the interleave un-jaggedify only apply on RESP3

add more RESP3 API change tests

compensate for XREAD having a different shape in RESP3

disable implicit RESP3 based on target server version

# Conflicts:
#	docs/Configuration.md
#	docs/ReleaseNotes.md
#	src/StackExchange.Redis/ConfigurationOptions.cs
#	src/StackExchange.Redis/ConnectionMultiplexer.cs
#	src/StackExchange.Redis/Interfaces/IConnectionMultiplexer.cs
#	src/StackExchange.Redis/ServerEndPoint.cs
#	tests/StackExchange.Redis.Tests/PubSubTests.cs
#	tests/StackExchange.Redis.Tests/TestBase.cs

…ugh DBs

…erver version and protocol

- add a few missing ForAwait
- add onnectWithTiming which is intended to find the netfx connect stall

- fix pubsub

- fix cluster tests (false positive: we *expect* zero subscriber connections in RESP3)
- fix flakey Lua test (unrelated to resp 3)
- separate the Me() key in protocol dependent tests

…rtion unless explicitly requested

…die)

- fix tests that assumed Create returned a real ConnectionMultiplexer

…t the fixture hold broken connections

…esting

This is a simpler approach to duplicating tests, but also fixes many - when changing things I noticed several base classes were incorrect so we silently dropped many tests in the PR inadvertently.

We have to get off AppVeyor and on to something else, testing against 3.1 in the mainline is kinda bananapants at this point.

These were colliding on RESP2/RESP3 but also using an entirely different system...let's simplify!