REGR: repr of stringified floats (e.g. "3.50") drop ending "0"

[topper-123]

Example:

>>> ser = pd.Series(["3.50"])
>>> ser
0    3.5  # wrong repr, should show 3.50
dtype: object
>>> ser[0]
'3.50'  # this is ok

[topper-123]

This is a regression since 1.1. The first bad PR was #36800.

cc @dsaxton