Fix conda forge installs

[MilesCranmer]

This will change the default behavior of PySRRegressor to not update the Julia packages when .fit is called. i.e., this will set update=False by default, rather than update=True. I think this is more in line with what a user would expect - that once they install PySR and run pysr.install(), all packages will remain static.

It is also causing trouble for conda-forge, since the pysr package actually includes all dependencies when first built - if any new dependency is needed, then pysr will break. This fixes #201. cc @ngam.

To update Julia packages, in pip or in conda, you may run pysr.install() again, or set update=True.

TODO:

• Create test for expected behavior when a user runs pip install -U pysr without running pysr.install() a second time.
• Mention in docstring for update kwarg that the user should make sure their pysr version is up-to-date first, before running install() - so there aren't any missing packages in the depot.
• Bump version number.
• Create check that the SymbolicRegression.jl version matches what PySR expects.

[MilesCranmer]

There isn't a clean way to check a Julia package's version right now, so I'll leave that for later. Just added a PACKAGE_VERSION constant to SymbolicRegression.jl, so that can be used in the future.