[BUG] Investigate regexp failures with unicode input

[andygrove]

Describe the bug
Investigate regexp issue discovered during testing #5450

The following test failed in CI but could not be reproduced locally. The issue appears to be specific to the superscript character ¹.

  test("issue 5450 ci failure") {
    val pattern = Seq("D?")
    val inputs = Seq("\u000E\u0011\u0006h¹TEST\ud720AA")
    assertCpuGpuMatchesRegexpReplace(pattern, inputs)
  }

Steps/Code to reproduce bug
See above.

Expected behavior
Behavior should be consistent between CPU and GPU or we should fall back to CPU.

Environment details (please complete the following information)
Failed in CI.

Additional context
None

[andygrove]

This should get fixed by #5627

[anthony-chang]

Fixed with #5776