Because it also won't match the sophisticated format we have for
detecting reserved names. We don't want to confuse the user with 2
error messages

If the cached html_field for a markdown_field is `nil` while the
mfarkdown_field is not, it needs to be refreshed.

Reject the part following `/*namespace_id/:project_id` for paths
containing 2 wildcard parameters

So we can block `objects` only when it is contained in `info/lfs` or `gitlab-lfs`

This reflects better that it validates paths instead of a namespace model

Since the namespacevalidator now knows the difference between a
top-level and another path, this could all be handled there.

The first part of a full path needs to be validated as a `top_level`
while the rest need to be validated as `wildcard`

`Project` uses `ProjectPathValidator` which is now a
`NamespaceValidator` that skips the format validation.

That way we're sure we are using the same collection of reserved
paths.

I updated the path constraints to reflect the changes: We now allow
some values that are only used on a top level namespace as a name for
a nested group/project.

These routes seem to be taken

This reverts merge request !10845

- Upgrade Gitaly-version