Speed up ECORR code using fastshermanmorrison

[abhisrkckl]

The package in question is https://github.com/nanograv/fastshermanmorrison written by @vhaasteren.

It implements the Sherman-Morrison lemma computation that's needed while computing the GLS chi2 when ECORRs are present. It's a major performance bottleneck in PINT right now. Since this package is written in Cython and a bit of straight-up C, it should offer significant speedup.

This package is pip and conda-installable according to its README file, so there is no issue with dependency management.

[vhaasteren]

I would have to work a bit to make fastshermanmorrison conda-installable. I did just figure out how to do a conda forge release for PTMCMCSampler, so I could make an effort. Right now it is not yet in conda-forge.

[dlakaplan]

For the moment PINT needs intel processors, so the ARM issue may not affect us here.

FWIW I did just install and test on a M1 Mac but with an Intel conda installation.