fix how int filter handles scientific notation

[flpm]

What was done:

• handle the case where a string passed to the int filter would cause an OverflowError when float() is called. E.g. '1e100000.
• add a new test case to test_int()
• add entry in CHANGES.rst

fixes #1921

[davidism]

git rebase --onto origin/3.1.x origin/main then git push -f