DOC: mention changes to max_rows behaviour in np.loadtxt

[charris]

Backport of #21854.

I think this breaking change should be documented. Also, an example on how to replicate previous behaviour regarding comment lines might be appropriate. But not sure where to include:

import itertools
lines = itertools.islice(open("file"), start, end)
# where comment lines starts with '#'
filtered_lines = itertools.dropwhile(lambda x: x[:1] == '#', lines)
result = np.loadtxt(filtered_lines, ...)

Related issue: #21852

• DOC: mention changes to max_rows behaviour

• Clarify how line counting works in max_rows