Check for imageinfo rather than missing or invalid

Because shared-repo images are reported as missing, do not check
for that status. And to make things simpler, use the same handling
for invalid ones as well, and just check to see if any imageinfo
metadata has been returned (that's all we're interested in, after
all).

This is a follow-up to GH #512.

Bug: T372956

Return early for cover images that are not found

BookProvider::getCover() was already returning for no pages
returned, so this switches to API formatversion 2 and also
checks for the `missing` or `invalid` parameter in the first/only returned
page.

Bug: T370257

Move image URL fixing to where the URL is used

The change I made in #513 fixed the URLs after they were used
to fetch the images! This moves that same logic to the location
in which it'll actually work. Sorry for the noise!

Bug: T354242

Handle unqualified image src attrs

Add the domain name to image src values that don't have one, to
support WikiHiero. Also move the protocol-relative handling out
of the PageParser so that it can go along with the new domain
name adding in one place.

Bug: T354242

Throw exception for rejected page requests (#511)

This returns to the previous behaviour of throwing a
WsExportException when a wiki page is not found.

Also changes the error message to call it a 'page' rather than
a 'book', becuase it could apply to a subpage or the root-level
page, and it isn't referring to the whole book.

Bug: T280329

Merge pull request #510 from wikimedia/transcluded-refs

Fix issue with cleaning transcluded reference links

Limit the concurrency when fetching pages

For a production test case, BookProvider::getPages() was opening 3492
concurrent connections to the Wikimedia servers. That exceeds the
default maxmimum file descriptor limit, and there is a risk that the
tool will be blocked by WMF sysadmins.

So, limit the concurrency to 10 connections.

Merge pull request #497 from wikimedia/counter

Uses a UNIX semaphore to limit concurrent ebook-convert calls

Merge pull request #496 from wikimedia/no-pagecount

Make only one API request for cover image info

Merge pull request #495 from wikimedia/cover-thumb-width

Add iiurlwidth API param when fetching cover image info