naturalsize() truncates non-integer byte counts below the base ("1 Bytes") #459

On current main:

>>> import humanize
>>> humanize.naturalsize(1.5)
'1 Bytes'
>>> humanize.naturalsize(1.9)
'1 Bytes'
>>> humanize.naturalsize(1500)
'1.5 kB'

From 1 kB upwards the fraction is kept using format, but below the base the value is truncated with int(). The singular check also compares the untruncated value with 1, so 1.5 is shown as "1 Bytes".

In #239 you mentioned that "1.1 Bytes" would be better than "1 Bytes" for non-integer input, and that tests for non-integer input were missing.

Proposal:

• Format non-integer values below the base with format, like larger units (e.g. "1.5 Bytes", and "1.5B" with gnu=True).
• Keep integer input unchanged ("300 Bytes", "1 Byte").
• Add tests for non-integer input.

Would you be happy with this? If so, we'd like to send a PR.