silx.io.fioh5: Improved handling of missing data entries

[tifuchs]

Enables the loading of fio files with missing data entries, that can not be converted into the target data type.
Mostly uses the default behaviour of numpy.genfromtxt.

For example: Entries like <no data> or nan in a FLOAT data column are automatically converted to numpy.nan by numpy.genfromtxt.

Not all missing values can be handled:
Missing boolean or integer values cannot be flagged as missing (i.e. nan) in a bool or integer array and thus result in a ValueError.

Closes #4155

[woutdenolf]

Fix CI failures

 >       val = val.decode().lower()
E       AttributeError: 'str' object has no attribute 'decode'

Could you also add a unit test with FIO data that has missingvalues?

[t20100]

Thanks for the PR!

As already mentioned, the continuous integration fails highlighting an bytes/str issue and it would also be great to have a test with nan and no data