Semicolons lead to incorrect form data in some entries

[Mindful]

Entries for words containing semicolons have bad form entries, apparently as a result of having their form/headword split at the location of the semicolon.

pv wiktionary.jsonl \
| jq 'select(.word == "too long; didn'\''t read").forms'

Gives

 {
    "form": "too long",
    "tags": [
      "canonical"
    ]
  },
  {
    "form": "didn't read",
    "tags": [
      "canonical"
    ]
  },
  {
    "form": "tl;dr",
    "tags": [
      "alternative"
    ]
  },
  {
    "form": "tl/dr",
    "tags": [
      "alternative"
    ]
  },

And a bunch more correct forms which I won't include because they aren't relevant.

tl;dr has similar problems:

  {
    "form": "tl",
    "tags": [
      "canonical"
    ]
  },
  {
    "form": "dr",
    "tags": [
      "canonical"
    ]
  },
  {
    "form": "tldr",
    "tags": [
      "alternative"
    ]
  },

As well as other entries with semicolons.

Seems like the issue might be related to logic splitting on semicolons here:

wiktextract/src/wiktextract/extractor/en/form_descriptions.py

Line 230 in f492ef9

+ ")?( or |[,;]+)"

but this is a suggestion from my local Codex instance and I don't know enough about the wiktextract codebase to verify it.