Reorganise unit tests to better match source layout

[hatemhelal]

This breaks up the growing test cases that were added in test_experimental into multiple files.

• removes some duplicated test cases which weren't obvious but were adding to the overall time.
• IPU specific tests are now in their own test file.
• updates to .pytest.ini will default to showing the 10 slowest test cases which can help flag any unexpected slow tests as more functionality is added to this module.

[hatemhelal]

A couple observations from the unittest logs

============================= slowest 10 durations =============================
13.21s call     test/test_integrals_ipu.py::test_eri
11.19s call     test/test_integrals_ipu.py::test_kinetic
9.56s call     test/test_integrals_ipu.py::test_nuclear
8.09s call     test/test_integrals.py::test_eri
6.58s call     test/test_integrals_ipu.py::test_overlap
4.50s call     test/test_integrals.py::test_nuclear
2.92s call     test/test_interop.py::test_gto
2.09s call     test/test_integrals.py::test_water_nuclear
1.93s call     test/test_integrals.py::test_kinetic
1.19s call     test/test_integrals.py::test_water_overlap[sto-3g]
======= 22 passed, 2 skipped, 1 xfailed, 3 warnings in [69](https://github.com/graphcore-research/pyscf-ipu/actions/runs/6458641631/job/17532758834?pr=122#step:7:70).21s (0:01:09) ========

The IPU tests use the IpuModel with 46 tiles -> experimenting with only 4 tiles doesn't noticeably speedup the tests

[awf]

Great to link to the logs from the PR too :)