Processing in wrap_float_literals() can't work line-by-line, as the
tokenize.unparse() doesn't preserve indentation in this case.  E.g.:

$ python -m mpmath --no-ipython
>>> def f():
...     return 1.2
...
Traceback (most recent call last):
  ...
  File "<unknown>", line 2
    return mpf('1.2')
    ^^^^^^
IndentationError: expected an indented block after function definition on line 1

While python/cpython#130662 being discussed, lets keep previous (more
strict) formatting specification.

Python docs doesn't specify 0-padding behavior, when alignment is
specified.  Lets be more strict here as Fraction/Decimal's: just reject
0 flag in this case.

float's behavior is just odd:
>>> format(1.123, '<020f')
'1.123000000000000000'