[raams95]

lei.validate() says it checks the length, format and check digits, but it only runs the ISO 7064 Mod 97, 10 check. So any string that happens to pass the checksum is accepted:

>>> from stdnum import lei
>>> lei.validate('1')
'1'
>>> lei.validate('0213800KUD8LAJWSQ9D15')   # valid LEI with a leading zero
'0213800KUD8LAJWSQ9D15'
>>> lei.validate('213800KUD8LAJWSQ9DAD')    # letters as check digits
'213800KUD8LAJWSQ9DAD'

An LEI (ISO 17442) is always 20 characters, letters and digits only, and the last two are the check digits. Mod 97, 10 always yields two decimal digits there. GLEIF describes the structure here: https://www.gleif.org/en/about-lei/iso-17442-the-lei-code-structure

This change makes validate() raise InvalidFormat for characters outside 0-9/A-Z, InvalidLength if the number isn't 20 characters, and InvalidFormat if the last two characters aren't digits. It doesn't touch compact() (the upper-casing discussed in #509 stays as it is). All the real LEIs in tests/test_lei.doctest still validate.

I added cases to tests/test_lei.doctest. They fail on master and pass with the change. The full pytest run passes with 100% coverage, and flake8 . is clean.