MMark is confused by pipes in code spans inside tables

[mrkkrp]

For example, this table is currently considered invalid, while it's perfectly fine:

Thing produced | Quotation syntax | Type
---------------|------------------|----------
Declaration    | `[d| … |]`       | `Q [Dec]`
Expression     | `[e| … |]`       | `Q Exp`
Type           | `[t| … |]`       | `Q Type`
Pattern        | `[p| … |]`       | `Q Pat`

[mrkkrp]

Also, we should be able to recover from parse errors in tables and continue parsing.

[mrkkrp]

Interestingly, GitHub's own markdown processor is also buggy and is fooled by pipes in code spans as well.

[mrkkrp]

Also, we need to figure out what to do when markdown looks like it could be interpreted as a list and as a table at the same time.