proposal: support docs on separate adapter

[costela]

This allows exposing the spec and docs on a separate adapter, potentially under a separate path and/or middleware stacks.

E.g.: (assuming internalRouter and mainRouter are two chi.Router)

opts := huma.DefaultConfig("foo", "0.0.1")
opts.DocsAdapter = humachi.NewAdapter(internalRouter)
opts.CreateHooks = []func(huma.Config) huma.Config{
	huma.DefaultSchemaLinkHook(huma.DefaultSchemaRefPrefix, internalPrefix), // assuming internalRouter above is mounted under internalPrefix
}

adapter := humachi.NewAdapter(mainRouter)

huma.NewAPI(opts, adapter)

This is a bit of a WIP:

• I'm not 100% sure the removal of getAPIPrefix call in the stoplight template is really as harmless as I think
• the other remaining getAPIPrefix call will currently break the schema links if the API is mounted somewhere other than the root.

But regardless of the limitations above: WDYT about the general idea?

[codecov]

Codecov Report

Attention: Patch coverage is 75.00000% with 7 lines in your changes missing coverage. Please review.

Project coverage is 93.06%. Comparing base (4977a7a) to head (dbc2946).

Files with missing lines	Patch %	Lines
transforms.go	33.33%	2 Missing and 2 partials ⚠️
api.go	75.00%	2 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #775      +/-   ##
==========================================
- Coverage   93.10%   93.06%   -0.04%     
==========================================
  Files          23       23              
  Lines        5296     5295       -1     
==========================================
- Hits         4931     4928       -3     
- Misses        313      314       +1     
- Partials       52       53       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[costela]

Actually, now that I look at the code, it looks like NewSchemaLinkTransformer's prefix parameter isn't even being used? 🙈

I've pushed a second take at this. The new one uses the first parameter to NewSchemaLinkTransformer, so it may be considered a breaking change?
Also I've tried to keep respecting the prefix detection via the OpenAPI.Servers, though I cannot find any place in the documentation where that's actually recommended? Maybe we can skip this backward compatibility and document explicitly how to mount stuff below the root path? There's a commented block in TestChiRouterPrefix that shows the two possible variants.