Before, we always tried to infer a file's format, even if it was
explicitly specified in the collection's config. This commit makes it so
that we always use the format from the config if it is specified, and
only if it is not set do we try to infer it from the file extension.
When entries were loaded for the editorial workflow dashboard, they were
all assumed to be FrontMatter/MarkDown files. This PR allows them to be
any supported format.
* Clean up frontmatter formatter.
* Move `formatToExtension`.
* Use plain objects for file formatters.
* Use same parsers for files and frontmatter.
We want to use our file parsers for frontmatter, instead of the builtin
ones, as they process some formats (images, dates) properly.
* Cleanup YAML frontmatter parser code.
