Extend `.parse_item_description`

[khaeru]

After the initial version added in #137 (per #133), looking at the Codelist=IMF:CL_AREA(1.15) specimen reveals some nuanced uses.

For each of these either improve the parser (if not too complex) or add an explicit statement in the docs (+ test) that such usage is not supported.

Also check:

• SDMX guidelines to see if these usages are mentioned.
• Item schemes from other data providers, to see if usage is similar or different.

Maybe support:

1. Multiple lines. Search for matching lines among all those in the description, last-to-first; assert there is only one parseable line.
2. Parenthetical comments:

   A1 - CA - CL (from 2010) - CO (from 2020) - MX - US

   These could be separated and returned as a third item in each tuple, e.g. (sub, "CL", "from 2010")

3. Expressions following a more narrative description and a ";" on the same line:

   W12=sum of (ISO 3166-1 alpha-2) partner countries; W12=W1-1A-W19

Probably too complex:

1. Parentheses:

   D5A = D5 - (BR + CA + CN + IN + JP + RU + CH +US)

2. Ellipses:

   5AA+5B+…+5Z+7A+…7M

3. Expressions in parentheses following text:

   Reference / counterpart area (W2.W0+W1.W2)