BUG: cannot set colorbar norm to SymLogNorm with base !=10

[neutrinoceros]

Bug report

Bug summary

Since yt 4.1.0, setting arbitrary colorbar norms is supported, however internally, isinstance(norm, SymLogNorm) is used with no regards to the actual base (base 10 is implicitly assumed) to trigger some hard coded special behaviours (most importantly setting minor ticks, which isn't handled well by matplotlib for this norm type).

Code for reproduction

import yt
from matplotlib.colors import SymLogNorm

yt.set_log_level("warning")

ds = yt.load_sample("IsolatedGalaxy")
p = yt.SlicePlot(ds, "z", ("gas", "velocity_x"))
p.set_norm(("gas", "velocity_x"), SymLogNorm(linthresh=1, base=2))
p.save("/tmp/test_sym.png")

Actual outcome

yt : [WARNING  ] 2023-05-21 14:59:45,100 Dropping norm constraints (norm_type=<class 'matplotlib.colors.Normalize'>)

Expected outcome

Ticks should be powers of 2 instead of powers of 10.

Version Information

• Operating System: MacOS
• Python Version: 3.11.1
• yt version: 4.1.x
• Other Libraries (if applicable): matplotlib 3.7.1

[neutrinoceros]

I find that while this is definitely fixable for matplotlib 3.5 and newer, it is much harder (dare I say, not worth it) for older versions, because SymLogNorm object do not expose the log base they are using, not even in private attributes.

I'll be working on a fix within the assumption that dropping support for Matplotlib 3.2, 3.3 and 3.4 is now acceptable (albeit an unresolved conversation, see #4379).