Markdown-it-py

This release is primarily to replace the `attrs` package dependency,
with the built-in Python `dataclasses` package.

This should not be a breaking change, for most use cases.

- ⬆️ UPGRADE: Drop support for EOL Python 3.6 (194)
- ♻️ REFACTOR: Move `Rule`/`Delimiter` classes from `attrs` to `dataclass` (211)
- ♻️ REFACTOR: Move `Token` class from `attrs` to `dataclass` (211)
- ‼️ Remove deprecated `NestedTokens` and `nest_tokens`
- ✨ NEW: Save ordered list numbering (192)
- 🐛 FIX: Combination of blockquotes, list and newlines causes `IndexError` (207)

- 🐛 FIX: Crash when file ends with empty blockquote line.
- ✨ NEW: Add `inline_definitions` option.
This option allows for `definition` token to be inserted into the token stream, at the point where the definition is located in the source text.
It is useful for cases where one wishes to capture a "loseless" syntax tree of the parsed Markdown (in conjunction with the `store_labels` option).

- ⬆️ Update: Sync with markdown-it v12.1.0 and CommonMark v0.30
- ♻️ REFACTOR: Port `mdurl` and `punycode` for URL normalisation (thanks to hukkin!).
This port fixes the outstanding CommonMark compliance tests.
- ♻️ REFACTOR: Remove `AttrDict`.
This is no longer used is core or mdit-py-plugins, instead standard dictionaries are used.
- 👌 IMPROVE: Use `__all__` to signal re-exports