Better document fill_value argument of read_*()

[alixdamman]

Reported by user GT.

He got nan instead of 0 although passed fill_value=0 to read_excel().

[gdementen]

I cannot reproduce this. I need a reproducer to be able to fix it.

FWIW, we have a test case for this at:
https://github.com/larray-project/larray/blob/master/larray/tests/test_array.py#L3402
https://github.com/larray-project/larray/blob/master/larray/tests/test_array.py#L3425
and the corresponding tests when using openpyxl:
https://github.com/larray-project/larray/blob/master/larray/tests/test_array.py#L3468

[gdementen]

(I am going to ask gt)

[gdementen]

In fact, there was no bug at all: just a misunderstanding of what fill_value does.

[gdementen]

will change the meaning of the argument in #1028 instead of changing the documentation