Fix parsec.validate list parser

[wxtim]

Investigation

Whilst working on https://github.com/cylc/cylc-flow/issues/5120 @MetRonnie and I noticed a number of untoward things about the functions strip_and_unquote_list and _unquoted_list_parse.

• Whilst test coverage is 100% I don't think that the tests represent the full range of sane input to these functions.
• The doctest for strip_and_unquote_list implies that (None, ' 1 , "2", 3') ⇒ ['1', '"2"', '3'] which seems wrong given the name of the function.
• A number of real world examples such as <foo=1, bar=2> are not tested at all.
• Some of the existing unit tests assert strange behavior 'a, b, c\n"d"' ⇒ ['a', 'b', 'd']

The functions may well be wrong, but users may depend on them being as they are so it's important to be careful with any changes.

[MetRonnie]

To add to this list...

Nonsensical difference between these two depending on whitespace:

>>> list(ParsecValidator._unquoted_list_parse(None, '1,"2",3'))
['1', '2', '3']

>>> list(ParsecValidator._unquoted_list_parse(None, '1, "2", 3'))
['1', '"2"', '3']

Unexpected splitting on </> chars if the first item in the list is quoted.

>>> ParsecValidator.strip_and_unquote_list(None, '"1", A<x>,43')
['1', 'A', '<', 'x', '>', '43']

And from #5120

>>> ParsecValidator.strip_and_unquote_list(None, '1, "2, 3", 4')
['1', '"2', '3"', '4']  # should be ['1', '2, 3', '4']

(Use case:)

[runtime]              
   [[I, J]]           
   [[A<x,y>]]         
   [[a1<x,y>, a2<x,y>, a3<x,y>]]
       inherit = I, "A<x,y>", J 

---

I suggest dropping the shlex branch of ParsecValidator.strip_and_unquote_list() https://github.com/cylc/cylc-flow/blob/f88f32e2bf14568664d27f9452c655baa45eb9b7/cylc/flow/parsec/validate.py#L554-L560

and then go about fixing these issues