BUG: repr aligning left for string dtype columns

[phofl]

• closes BUG: wrong alignment of column names in the DataFrame repr when using pyarrow-backed string dtype #54797 (Replace xxxx with the GitHub issue number)
• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added type annotations to new arguments/methods/functions.
• Added an entry in the latest doc/source/whatsnew/vX.X.X.rst file if fixing a bug or adding a new feature.

cc @jorisvandenbossche this is a bit hacky, but changing format_array breaks index-related stuff. Don't feel comfortable to do this now, can be a follow up (we are casting to object anyway, so this does not impact performance)

[jorisvandenbossche]

Yes, if the formatter converts to object dtype anyway, then this looks like a fine short term solution.

[jorisvandenbossche]

Suggested change

	if is_string_dtype(values.dtype):
	values = np.asarray(values)
	if is_string_dtype(values.dtype):
	# ensure we have an object dtype numpy array
	values = np.asarray(values)

[jorisvandenbossche]

this is a bit hacky, but changing format_array breaks index-related stuff.

Looking at the code in Index._format_with_header, it seems we just never use format_array in case of object dtype (i.e. current strings). And in the code path for non-object dtype calling format_array, we specify justify="left". That's the reason for this behaviour, and so something we would have to change (or let the value depend on the dtype; I don't fully understand the reasoning for this justify="left" for other dtypes)

[phofl]

Yeah that's correct, we can set justify=all, but this screws with Index formatting if the index has a NaN because we are doing something weird. But that's too big for this here, that's why I went with the short term solution

[jorisvandenbossche]

I don't fully understand the reasoning for this justify="left" for other dtypes

You can get this "bug" as well for other data types because of this left alignment. For example with numeric index:

In [14]: df = pd.DataFrame({99999999999999999: [1, 2, 3], 1: [4, 5, 6]})

In [15]: df
   99999999999999999  1                
0                  1                  4
1                  2                  5
2                  3                  6

(will move this to the issue)