[join-] prevent diff error for empty merge cells

[midichef]

After merging sheets, cells in MergeColumn hold None for rows that were not present on the sheet that the MergeColumn came from. Visidata shows errors when those ⌀ cells appear on screen:

test_merge.vdj.txt
cd sample_data; vd -p test_merge.vdj

File "/home/midichef/.pyenv/versions/3.12.2/lib/python3.12/site-packages/visidata/features/join.py", line 179, in <lambda>
    CellColorizer(0, 'color_diff', lambda s,c,r,v: c and r and isinstance(c, MergeColumn) and c.isDiff(r, v.value))
File "/home/midichef/.pyenv/versions/3.12.2/lib/python3.12/site-packages/visidata/features/join.py", line 170, in isDiff
    return col and value != col.getValue(row[col.sheet])
...
File "/home/midichef/.pyenv/versions/3.12.2/lib/python3.12/site-packages/visidata/utils.py", line 134, in getitemdeep
    return obj[k]
TypeError: 'NoneType' object is not subscriptable

The return statements in this PR could be shortened to a one-line conditional: return col and ((row[col.sheet] is None) or (value != col.getValue(row[col.sheet]). But I broke it up to multiple lines because I found row and col to be quite difficult objects to think about. In fact, I wanted to add a comment to describe what situation would make row[col.sheet] be None. But row comes from groupRowsByKey(), and it's complex enough that I'm not sure I can describe it correctly:

visidata/visidata/features/join.py

Line 121 in a48c4ee

rowsByKey[key] = [

This change to isDiff() only silences the error. It doesn't actually change the color of the ⌀ cells, for two reasons:

1. color_diff is not set by default, so isDiff() has no visible effect.
2. Even if color_diff is set, it would not be used for ⌀, because color_note_type takes precedence.

[saulpw]

This is good, thanks for fixing this and expanding isDiff().

[saulpw]

I found row and col to be quite difficult objects to think about. In fact, I wanted to add a comment to describe what situation would make row[col.sheet] be None

groupRowsByKey() can certainly be confusing, but this should be pretty straightforward here. Each row on the joined sheet is a dict, with the key being the source sheet, and the value being the source row from that source sheet. So row[col.sheet] is None when there is no matching row from that sheet.