14: feat: add the work pages

[juliantellez]

What. Seven pages: a work index and one page per engagement, for ITV, castLabs, Sky, DAZN,
Utility Warehouse and Plentific. Wired into the header, the footer and the home page.

They are not conventional case studies, on purpose. A conventional one leads with the client's
numbers, and those are the client's to publish. So each page leads with the shape of the problem
instead: what goes wrong in that kind of platform, how it is found, what the fix looks like. The
client is named as where the work happened, with the sector and the period, and nothing about their
systems, metrics or roadmap appears.

This is weaker than a case study with a number in it, and worth saying plainly. 80 per cent of
business buyers use case studies and the absence of a specific outcome is a known disqualifier. The
version that would convert better needs one client to agree to be quoted. Until then this is the
honest form, and it still demonstrates the expertise, because the description of the problem is
itself evidence of having lived it.

Structured data. Article and BreadcrumbList per page. Each links to the services the work
touches, and each service is reachable from the footer of every page.

Work is back in the header, now that there is a page behind it. It pointed at /services
before, which is why I removed it.

Tests. 51, up from 40. Seven added: every page answers 200, one h1, a canonical pointing at
itself, the client named, the service cross links resolve, Article structured data parses, and no
client metric appears on any of them.

Three were mutation checked. Putting a metric back into the ITV copy fails the confidentiality
test. Pointing an engagement at a service that does not exist fails the cross link test. Both went
red then green again.

Closes #14. Part of #2.