Improve arraytranslateEvaluator for Power

[knn-k]

This commit removes the code generation for checking length == 0 in arraytranslateEvaluator for Power.
The helper functions for arraytranslate are responsible for checking the case of length == 0.

[knn-k]

The arraytranslate evaluator generates code like this without this change:

    0x7a940ffc016c 000000f8 [    0x7a940f0d95f0]                   10   Label L0097:    ; (Start of internal control flow)
    0x7a940ffc016c 000000f8 [    0x7a940f0d9680] 2c250000          10   cmpdi   cr0, gr5, 0
    0x7a940ffc0170 000000fc [    0x7a940f0d9720] 4181000c          10   bgt     cr0, Label L0098
    0x7a940ffc0174 00000100 [    0x7a940f0d97c0] 38600000          10   li      gr3, 0000000000000000
    0x7a940ffc0178 00000104 [    0x7a940f0d9860] 48000018          10   b       Label L0099
    0x7a940ffc017c 00000108 [    0x7a940f0d98f0]                   10   Label L0098:
    0x7a940ffc017c 00000108 [    0x7a940f0dc9a0] 60600000          10   ori     gr0, gr3, 0x0
    0x7a940ffc0180 0000010c [    0x7a940f0d9a10] 481ffba1          10   bl      00007A94101BFD20 ; Direct Call "__arrayTranslateTROT255 (ISO8859_1)"
    0x7a940ffc0184 00000110 [    0x7a940f0dc900] 7c630278          10   xor     gr3, gr3, gr0
    0x7a940ffc0188 00000114 [    0x7a940f0dc860] 7c600278          10   xor     gr0, gr3, gr0
    0x7a940ffc018c 00000118 [    0x7a940f0dc720] 60030000          10   ori     gr3, gr0, 0x0
    0x7a940ffc0190 0000011c [    0x7a940f0d9b10]                   10   Label L0099:    ; (End of internal control flow)

This commit removes the cmpi and the bgt from the main line of the generated code, in addition to eliminating the inappropriate register dependency at labelArrayTranslateDone that caused annoying register shuffles.
All the arraytranslate helper functions, except for TR_PPCarrayTranslateTRTO, work as expected when the length is zero. A PR in OpenJ9 changes the TRTO helper to handle the case of length zero.

The performance degrades with this change when the length is zero, but I assume that does not happen very often.

[knn-k]

Jenkins build plinux,aix

[knn-k]

There is no need for a coordinated merge. eclipse-openj9/openj9#21087 is to be merged first, and this PR to follow.

The test failure on plinux is a known problem: Issue #6571:

[2025-02-10T02:18:02.862Z] 31: [0;32m[----------] [m16 tests from Special/PPCDirectEncodingTest
[2025-02-10T02:18:02.862Z] 31: free(): invalid next size (normal)

[knn-k]

@zl-wang FYI.

[zl-wang]

LGTM