Delete an unnecessary pessimization for x86.

[sandreenko]

Now if we have an incoming struct its fields can be enregistered on x86.
It includes both cases:

1. the struct is coming on stack but we copy it to regs;
2. the struct is coming on reg, in x86 case it is only possible for something like struct {int} and the field is used from the reg.

benchmarks.run.windows.x86
  Total bytes of delta: -2195 (-1.25% of base)
  440 total files with Code Size differences (340 improved, 100 regressed), 120 unchanged.
libraries.crossgen.windows.x86
  Total bytes of delta: -10702 (-1.71% of base)
  2274 total files with Code Size differences (1827 improved, 447 regressed), 1043 unchanged.
libraries.crossgen2.windows.x86
  Total bytes of delta: -5375 (-3.53% of base)
  938 total files with Code Size differences (771 improved, 167 regressed), 169 unchanged.
libraries.pmi.windows.x86
  Total bytes of delta: -18612 (-1.50% of base)
  3752 total files with Code Size differences (2828 improved, 924 regressed), 1593 unchanged.
tests.pmi.windows.x86
  Total bytes of delta: -3563 (-6.94% of base)
  419 total files with Code Size differences (380 improved, 39 regressed), 59 unchanged.
tests_libraries.pmi.windows.x86
  Total bytes of delta: -3892 (-0.61% of base)
  1220 total files with Code Size differences (890 improved, 330 regressed), 1705 unchanged.

the regressions are from additional reg pressure + CSE, nothing stands up.
libraries diffs are very nice, hope some people are still using x86 :-)

I will do it for arm32 as well, this condition was breaking a struct contract that I need to satisfy for struct enreg, the contract is:
if struct's promotion type is 'DEPENDENT' the struct should be marked as 'doNotEnreg`.

[sandreenko]

/azp run runtime-coreclr jitstress

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[sandreenko]

A diff example:

base:
2c415d0: 50                     push    eax
2c415d1: 89 14 24               mov     dword ptr [esp], edx
2c415d4: 8b 04 24               mov     eax, dword ptr [esp]
2c415d7: 89 01                  mov     dword ptr [ecx], eax
2c415d9: 59                     pop     ecx
2c415da: c3                     ret

diff:
2c41730: 89 11                  mov     dword ptr [ecx], edx
2c41732: c3                     ret

[sandreenko]

The failures are due to #52988 (unrelated).

[sandreenko]

PTAL @dotnet/jit-contrib

[AndyAyersMS]

LGTM