[mono][jit] Pass generic context when ldftn of generic method

[BrzVlad]

Fixes cases like mono/mono#20962

[lambdageek]

Also fixes #66337 (which is the same as mono/mono#20962)

[lambdageek]

/azp run runtime-extra-platforms

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[lambdageek]

we need something for llvm, too, I think:

runtime/src/mono/mono/mini/jit-icalls.c

Lines 50 to 52 in 906146a

	if (mono_method_needs_static_rgctx_invoke (method, FALSE))
	/* The caller doesn't pass it */
	g_assert_not_reached ();

[SamMonoRT]

Contributes to #63548 via #49904

[vargaz]

Looks ok, but the last time I tried this in mono, it broke some other tests.

[BrzVlad]

Looks ok, but the last time I tried this in mono, it broke some other tests.

The original change was https://github.com/mono/mono/pull/20969/files. That change assumed that it needs to pass a rgctx for every indirect call. The problem with it is that it is not true in the case of tailcalls, where the called method is already known and the rgctx is passed as part of the call. OP_TAILCALL uses a MONO_PATCH_INFO_METHOD_JUMP.

Also the jump trampolines used as part of ldftn are not optimal, since no patching is taking place. They always go into the common_call_trampoline. We will probably need something else there at some point.

[SamMonoRT]

/azp run runtime-extra-platforms

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[SamMonoRT]

azdo reports an all clean build : https://dev.azure.com/dnceng/public/_build/results?buildId=1668431&view=results