gateway: tests for simpler superfluous namespace handling

[hacdias]

Closes #344. See comments below.

Kubo: ipfs/kubo#10322

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (89bceff) 65.48% compared to head (a393006) 65.60%.

@@            Coverage Diff             @@
##             main     #572      +/-   ##
==========================================
+ Coverage   65.48%   65.60%   +0.12%     
==========================================
  Files         207      207              
  Lines       25605    25597       -8     
==========================================
+ Hits        16768    16794      +26     
+ Misses       7361     7324      -37     
- Partials     1476     1479       +3     

Files	Coverage Δ
gateway/handler.go	78.27% <100.00%> (+1.81%)	⬆️

... and 14 files with indirect coverage changes

[lidel]

I did not look too deep, but I think this breaks valid content paths like /ipfs/ipfs/file.txt in places like subdomains – see the failing test of {{scheme}}://{{cid}}.ipfs.{{host}}/ipfs/ipns/ at https://github.com/ipfs/gateway-conformance/blob/193833b91f2e9b17daf45c84afaeeae61d9d7c7e/tests/subdomain_gateway_ipfs_test.go#L244-L259

Not sure how feasible, but it may be easier to move handleSuperfluousNamespace check from very front of the request processign pipeline, to very end: trigger it on "not found" code path (e.g. the place where we look for _redirects or produce HTTP 404 response)

[hacdias]

@lidel I have actually realised #344 is actually not an issue. When I initially tested, they didn't redirect because I hadn't set the Paths field in the configuration 🤦

Nevertheless, we can use this PR as a simplification and also a bunch of tests.

[hacdias]

@lidel it seems that we tested this in Kubo too. See here: ipfs/kubo#10322 too.

[lidel]

lgtm, thank you for cleaning this up!