generate: Add power decay spatial profile

[cwindolf]

This PR adds another spatial decay family to the template simulator. Instead of just exponential, there's also a power law type decay.

I'd like to switch over totally to SpikeInterface for my simulation studies, and with this and the temporal correlations thing I think that should be enough for me.

[cwindolf]

I think the test fails are unrelated?

[alejoe91]

I think the test fails are unrelated?

yes very unrelated. should be fixed by #4490

[chrishalcrow]

Yeah, weird test failure. The complete tests should not be trying to include .github/scripts/test_kilosort4_ci.py. I'll look into it...

[cwindolf]

The test fail seems to happen because in "src/spikeinterface/preprocessing/tests/test_pipeline.py::test_loading_provenance", there is a place where bad channels are detected and discarded; this discards all the channels and there is some file-seek thing where .seek(0 - 1) is called because the data size is 0.

It passes locally for me, but it seems like my 2 PRs are the only ones where this test fails? I am a bit confused.

[cwindolf]

Got it figured out. The failing test was flaky because detect_bad_channels() can be touchy. It strongly depended on the parameters of the 10 units generated in that test, and I had changed the order of random numbers being generated by putting my new unit param in the middle of the defaults dict which is traversed in order. I just moved it to the end to make the RNG path the same, but didn't fix the flaky test (which is fine, it's just there to test loading provenance stuff).