Kill correct registers for VLM

[Spencer-Comin]

Account for vector registers when calculating real register to check if spill is needed after a load multiple instruction. This fixes the behaviour of uneccessarily and incorrectly spilling GPRs when using VLM. An example of the error from my experience trying to accelerate crc32c on Z:

[     0x3ff8dcbd580]                          VLM     VRF_0095,VRF_0098,#415 0(&GPR_0033)VRF_0095(0/6)~FPR5 VRF_0098(0/6)~FPR8 
details:                      trying to free GPR_0099 for killed reg GPR6 by loadmultiple 

details:                      BEST FREE REG for GPR_0099 is GPR14 

 [     0x3ff8ddb3610]                          LGR     GPR6,GPR14       details:                      trying to free GPR_0050 for killed reg GPR7 by loadmultiple 

details:                      BEST FREE REG for GPR_0050 is GPR9

 [     0x3ff8ddb37f0]                          LGR     GPR7,GPR9      
 [     0x3ff8dcbd580]                          VLM     VRF5,VRF8,#415 0(GPR2)

Here GPR7 and GPR6 are unnecessarily spilled since GPR5-GPR8 are incorrectly used in the check at [1] rather than VRF5-VRF8. This patch solves this problem.

[1] https://github.com/eclipse/omr/blob/f99c9dd2248e6ee35e40536b3c243eee33924407/compiler/z/codegen/OMRInstruction.cpp#L722

[Spencer-Comin]

@r30shah @joransiu

[r30shah]

Jenkins build zos,zlinux

[joransiu]

LGTM

[r30shah]

Change Looks Good to Me.
@0xdaryl As z/OS and Linux on Z builds have passed and Spencer has done testing through personal builds as well for this change, Can I request you to merge this change?

[r30shah]

Relaunching build with the rebased changes to make sure nothing breaks

[r30shah]

Jenkins build zos,zlinux

[r30shah]

@0xdaryl Can you please review/merge this change?