☂️ Transition client-side logic over to SingleInvokeInitiator & cluster objects

[mrjerryjohns]

Problem

Once #8894 has been fixed and SingleInvokeInitiator has been implemented, the client-side code that currently interfaces with CommandHandler should be ported over to using this new class + cluster objects.

However, to make this a gradual transition, it has been broken down into 5 different phases that are designed to maximize parallelism and ensure smaller sized PRs:

• C1: Fix-up CHIPClusters.cpp to internally use SingleInvokeInitiator #8958
• C2: Delete gCallbackMgr, CHIPClientCallbacks.cpp and CommandSender logic. #8960
• C3: Create parallel methods in CHIPClusters.cpp that accept generated cluster object #8961
• C4: Transition all command call-sites that use CHIPClusters.cpp over to cluster object-based method #8967
• C5: Clean-up un-used command methods in CHIPClusters.cpp #8968

This is an 'umbrella' issue that to be deemed fixed, requires the above 5 issues to be fixed first.

[mrjerryjohns]

Will add links to the above issues tomorrow...

[mrjerryjohns]

@bzbarsky-apple FYI.

[woody-apple]

What is the impact of this?

[bzbarsky-apple]

The main impacts I know about:

• Broken error-handling in commissioning because the APIs it's using don't propagate out errors correctly.
• Lots of nearly-unused code floating around that we could delete if we stopped using the last few bits of it.

Fixing #13371 is likely to fix this, fwiw.

[yunhanw-google]

this has been fixed for C1-C5