Pipeline improvements

[Anarh2404]

Proposed Changes

This PR based on awesome @stebet work in #706
I made some improvements and achieved even greater reduction in memory consumption and increased performance.
Here is what I did:

• Reduce the number of async paths (which generates AsyncStateMachine) in pipeline reader
• Improve binary readers/writers

All tests run, except for the APIApproval

Types of Changes

• Bug fix (non-breaking change which fixes issue #NNNN)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause an observable behavior change in existing systems)
• Documentation improvements (corrections, new content, etc)
• Cosmetic change (whitespace, formatting, etc)

Checklist

• I have read the CONTRIBUTING.md document
• I have signed the CA (see https://cla.pivotal.io/sign/rabbitmq)
• All tests pass locally with my changes
• I have added tests that prove my fix is effective or that my feature works
• I have added necessary documentation (if appropriate)
• Any dependent changes have been merged and published in related repositories

Further Comments

I added netstandard2.1 and few #if derectives, because some perf critical API is only available in the netstandard2.1/netcoreapp3.0. For netstandard2.0 and net461 added fallback to a less optimal implementation.

Memory test

512byte payload
#706

This PR

Simple perfomance test on my home pc

1 publisher, 1 consumer, no ack
Nuget 5.1.2

#706

This PR

1 publisher, 1 consumer, manual ack, 64 prefetch count
Nuget 5.1.2

#706

This PR

[michaelklishin]

Thank you. Please remove the test program. Publishing it in a separate repo under your account should be sufficient.

[michaelklishin]

Please rebase this against master.

Out of curiosity, do most of the gains come from the switch to BinaryBufferReader? Or the reduced number of pipeline paths

[lukebakken]

@Anarh2404 thanks for this pull request.

We can't merge this pull request as-is into master. You should open your pull request against the allocationImprovements branch in the https://github.com/stebet/rabbitmq-dotnet-client.git` fork of this repository.

Or, @stebet - would you mind reviewing these changes and, if acceptable, merge them into your stebet:allocationImprovements fork & branch? Thanks.

[lukebakken]

@stebet - and by "these changes" I mean the changes in @Anarh2404 's commit: 3147ab5

[stebet]

I'll take a look tomorrow :)

[Anarh2404]

@michaelklishin BinaryBufferReader has reduced memory consumption, but not only it. Reduction of async paths increased throughput.

@lukebakken @stebet I think it would be nice if this PR would be merged into a stebet:allocationImprovements, of course, if this acceptable.

[Anarh2404]

Made a test branch where i merged this PR and #732 :)

[Anarh2404]

@michaelklishin test program from the #706

[lukebakken]

I am assuming that the changes in this PR will be investigated for version 7.0.0