Is pymc.Gumbel the extreme maximum or minimum distribution?

It is the regular Gumbel distribution, suitable for modelling maxima. To verify this you can examine the logp and logcdf definitions in
pymc.distributions.continuous — PyMC 5.9.0 documentation

and make comparison versus equations for Gumbel distribution e.g. on Wikipedia:
Gumbel distribution - Wikipedia

Wikipedia gives a simple tip for how to model minima using the Gumbel distribution:

To model the minimum value, use the negative of the original values.

Hope this helps!