Percent-encoding parts of URL leads to a 404

Bug description

URLs with percent-encoded characters are not resolved properly. They lead to a 404 page even though the canonical page exists.

How to reproduce

1. Find a working GitBook URL, e.g. https://gitbook.com/docs/getting-started/quickstart
2. Percent encode one or more characters in the URL path, e.g. https://gitbook.com/docs/getting-started/quickstar%74
3. Notice that the page gives a 404 "Page not found" error, and that your browser has correctly folded the URL in the address bar to the canonical form (but it still requests the %-encoded form from the server!)

Additional context

Docs: https://developer.mozilla.org/en-US/docs/Glossary/Percent-encoding

Although it's strange to percent-encode non-reserved characters, it's still part of the standard. I found it useful for making links to pages with - characters, as they break automatic URL creation in Microsoft Word references, while %2d works as expected on most websites.

Start by reproducing the issue with the canonical GitBook URL and the percent-encoded path shown in the report, then compare the two requests. Use the linked MDN percent-encoding reference to confirm the expected behavior. Done means encoded non-reserved characters resolve to the existing canonical page instead of returning a 404.