Handle (mask & ignore) off-domain values natively in log plots

[sadielbartholomew]

At the moment, cf-plot supports logarithmic level plotting, i.e. when the colour scale is on a log scale to visualise the underlying data in a logarithmic fashion (not to be confused with log axes for the projected flat representation via xlog or ylog, etc.). However, when there are array values which are invalid for the domain of a log operation, namely zero or negative, they cause the plot to show an opaque error of ValueError: zero-size array to reduction operation maximum which has no identity. A log plot can still be produced by masking the invalid values out and then plotting with blockfill=True to handle the masked values (see example below), but this all needs to be done manually as prep. We should allow the log call to work as-is where the invalid domain values are masked automatically and the plot is converted to a blockfill, with a warning-level log message to indicate why this has occured.

MRE of failing naive log attempt and workaround

(From a real user support case today.)

>>> a = cf.read("liap_MMR_2006bco.nc")[0]
>>> >>> f = a[0, 0]
>>> f
<CF Field: long_name=actual contrail coverage (not area weighted)(long_name=hours since 1900-01-01 00:00 UTC at the first of the month(1), long_name=pressure_level(1), long_name=latitude(73), long_name=longitude(144)) %>
>>> f.shape
(1, 1, 73, 144)
>>> cfp.con(f)  # non-log plot, works
>>>
>>> # Attempt to get a log plot via the supported approach 
>>> cfp.setvars(level_spacing="log")
>>> cfp.con(f)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/slb93/git-repos/cf-plot/cfplot/cfplot.py", line 3438, in con
    clevs, mult, fmult = calculate_levels(
                         ^^^^^^^^^^^^^^^^^
  File "/home/slb93/git-repos/cf-plot/cfplot/cfplot.py", line 10183, in calculate_levels
    close_below = np.max(field[pts])
                  ^^^^^^^^^^^^^^^^^^
  File "/home/slb93/miniconda3/envs/old-sphinx-cf-doc-build-only/lib/python3.11/site-packages/numpy/_core/fromnumeric.py", line 3164, in max
    return _wrapreduction(a, np.maximum, 'max', axis, None, out,
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/slb93/miniconda3/envs/old-sphinx-cf-doc-build-only/lib/python3.11/site-packages/numpy/_core/fromnumeric.py", line 86, in _wrapreduction
    return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: zero-size array to reduction operation maximum which has no identity
>>> cfp.setvars(level_spacing="loglike")
>>> cfp.con(f)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/slb93/git-repos/cf-plot/cfplot/cfplot.py", line 3438, in con
    clevs, mult, fmult = calculate_levels(
                         ^^^^^^^^^^^^^^^^^
  File "/home/slb93/git-repos/cf-plot/cfplot/cfplot.py", line 10183, in calculate_levels
    close_below = np.max(field[pts])
                  ^^^^^^^^^^^^^^^^^^
  File "/home/slb93/miniconda3/envs/old-sphinx-cf-doc-build-only/lib/python3.11/site-packages/numpy/_core/fromnumeric.py", line 3164, in max
    return _wrapreduction(a, np.maximum, 'max', axis, None, out,
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/slb93/miniconda3/envs/old-sphinx-cf-doc-build-only/lib/python3.11/site-packages/numpy/_core/fromnumeric.py", line 86, in _wrapreduction
    return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: zero-size array to reduction operation maximum which has no identity
>>>
>>> # Workaround by masking values out of the valid domain then plotting with blockfill=True
>>> a_with_zeros_masked = a.where(cf.isclose(0.0), cf.masked)  # mask out all zero values - note you need cf.isclose() to deal with float precision differences
>>> cfp.setvars()  # reset level spacing from before
>>> cfp.con(a_with_zeros_masked[0, 0], blockfill=True)  # standard, non-log plot
>>> cfp.setvars(level_spacing="log")
>>> cfp.con(a_with_zeros_masked[0, 0], blockfill=True)   # logarithmic scale plot, works