* fix: non leading-tabs in markdown content (#1559)
Only replaces tabs at the beginning of a block construct. Tabs in the
middle of the item are unaffected.
All tests passing. Tabs in both GFM and CommonMark at 100%
fixes#1559
* update new/html_comments.html to preserve tab
* combine redundant if condition
* add test for tab immediately after blockquote character
* fix: rename inconsistent function
* fix: newline with 1~4 spaces below table
* fix: format html test
* fix: specify spaces only
* fix: specify spaces only
* fix: detect tab
* fix: detect tab
* fix: newline at the end of html
* fix: gfm table requires leading empty line
* test(gfm/table): add some test cases
* revert changes to `gfm.0.29.json`
* test(gfm/table): add `table_following_text` testcase
BREAKING CHANGE:
- Default export removed. Use `import { marked } from 'marked'` or `const { marked } = require('marked')` instead.
- `/lib/marked.js` removed. Use `/marked.min.js` in script tag instead.
- When using marked in a script tag use `marked.parse(...)` instead of `marked(...)`
BREAKING CHANGES:
- Tokenizers will create their own tokens with `this.lexer.inline(text, tokens)`. The `inline` function will queue the token creation until after all block tokens are rendered.
- `nptable` tokenizer is removed and merged with `table` tokenizer.
- Extensions tokenizer `this` object will include the `lexer` as a property. `this.inlineTokens` becomes `this.lexer.inline`.
- Extensions parser `this` object will include the `parser` as a property. `this.parseInline` becomes `this.parser.parseInline`.
- `tag` and `inlineText` tokenizer function signatures have changed.
Cannot use **(WhiteSpace)(WhiteSpace)(Return)** to break line while only one characher before it.
And if use special character, it will reset the character count
brought by [#1460](https://github.com/markedjs/marked/pull/1460)
Beginning and End delimiters for EM and Strong must be searched in a separate regex to work without lookbehinds. This invalidates the regex that skips over blocks (code, html, etc.) that take precedence over EM or Strong blocks.
Getting around this means we must now mask not only reflinks, but all enclosed blocks which were previously just skipped over in the Regex.
Add one check for overlapping Strong block when testing EM, now passes Commonmark 390 and 471
Reflinks are replaced with the string 'aaaa' of equal length to hide any asterisks or underscores that might have been inside, since these should not be included. Then, the resulting string is matched to the em regex. The positions of the captures are then used on the original string.
Added a check for the previous character to the *em* Tokenizer. Needed to pass any tests where the em block starts with a punctuation character (e.g. commonmark example 368)
