fix: resolve 404/500 error when downloading files containing hash in filename (#15799)

### What?

Fixes file downloads returning 404 for filenames containing `#` when
served through the REST API.

 ### Why?

Next.js catch-all `[...slug]` params decodes URL-encoded segments before
passing them for further handling. So a request for
`/api/media/file/document%20%23123.pdf` arrives with slug segments
`['media', 'file', 'document #123.pdf']`.

These decoded segments are joined and passed to handleEndpoints, but the
implementation of handleEndpoints is using `path-to-regexp` with a
configuration that expectes urlencoded path segments.

### How?

Re-encode each slug segment with `encodeURIComponent` before joining:

handleEndpoints already uses match(endpoint.path, { decode:
decodeURIComponent }), so params are decoded back to their original
values after matching

Also adds a test that creates a file with # in the filename and verifies
it can be served via REST.

Fixes #15798

---------

Co-authored-by: Paul Popus <paul@payloadcms.com>