Add StableHLO complex log to stablehlo-complex-math-expander pass

[pearu]

As in the title. This PR is created on top of the branch of #2679.

This PR improves the accuracy of JAX complex log function as follows:

Before
------
test_accuracy.py::test_unary[log-jax-cpu-complex64-default] maximal ULP difference: 4294967296
ULP difference == 0: 1612359
ULP difference == 1: 487617
ULP difference == 2: 306
ULP difference == 3: 140
ULP difference == 4: 68
ULP difference == 5: 49
ULP difference == 6: 26
ULP difference == 7: 35
ULP difference == 8: 19
ULP difference == 9: 15
ULP difference == 10: 15
ULP difference >= 11: 151

test_accuracy.py::test_unary[log-jax-cuda-complex64-default] maximal ULP difference: 4294967296
ULP difference == 0: 1797796
ULP difference == 1: 301398
ULP difference == 2: 1044
ULP difference == 3: 169
ULP difference == 4: 85
ULP difference == 5: 51
ULP difference == 6: 32
ULP difference == 7: 20
ULP difference == 8: 7
ULP difference == 9: 13
ULP difference == 10: 8
ULP difference >= 11: 177

After
-----
test_accuracy.py::test_unary[log-jax-cpu-complex64-default] maximal ULP difference: 3
ULP difference == 0: 1581126
ULP difference == 1: 519652
ULP difference == 2: 19
ULP difference == 3: 3

test_accuracy.py::test_unary[log-jax-cuda-complex64-default] maximal ULP difference: 2
ULP difference == 0: 1775914
ULP difference == 1: 324185
ULP difference == 2: 701

The corresponding accuracy patterns are available in pearu/functional_algorithms#46 (comment) .

[pearu]

@GleasonK , please review.

[GleasonK]

Hey! The team is mostly OOO until ~Jan 3-6th

cc @sdasgup3 @ghpvnist @abhigunj if any of you are back before the 6th PTAL

[ghpvnist]

LGTM. Once sqrt PR is merged, let's rebase and submit.