feat: support HTTP 429 with Retry-After

[hacdias]

This closes #188.

• Moved all - I think - logic related to errors to errors.go. I also added some simply sanity tests for this new error as it is its own custom type to support the RetryAfter field.
• Added gateway.ErrTooManyRequests. Returning an error of this type will make the gateway return 429. If the field RetryAfter above 0, it will set the Retry-After header. Also added a more "e2e" test for this, albeit arguable unnecessary since I added tests for webError.

[codecov]

Codecov Report

Merging #194 (1086d9d) into main (00e0249) will increase coverage by 0.25%.
The diff coverage is 77.06%.

@@            Coverage Diff             @@
##             main     #194      +/-   ##
==========================================
+ Coverage   29.75%   30.01%   +0.25%     
==========================================
  Files         100      101       +1     
  Lines       11268    11312      +44     
==========================================
+ Hits         3353     3395      +42     
- Misses       7547     7549       +2     
  Partials      368      368              

Impacted Files	Coverage Δ
bitswap/client/internal/session/session.go	0.00% <0.00%> (ø)
...tswap/client/internal/session/sessionwantsender.go	0.00% <0.00%> (ø)
bitswap/message/pb/cid.go	50.00% <0.00%> (ø)
bitswap/wantlist/forward.go	0.00% <ø> (ø)
files/serialfile.go	66.32% <0.00%> (ø)
gateway/handler.go	64.23% <50.00%> (-0.80%)	⬇️
gateway/errors.go	84.94% <84.94%> (ø)
files/multifilereader.go	81.81% <100.00%> (-0.21%)	⬇️

[lidel]

💭 something I've realized during this review is that Retry-After is not limited to 429 Too Many Requests, which is used when user is sending too many requests. There are cases where Retry-After is returned with 503 Service Unavailable (when it is not specific to user or their IP, could be global outage).

Pushed two commits which replace all standalone errors with generic wrappers: ErrorResponse and ErrorRetryAfter, and attempted to wire that up in ipfs-inactive/bifrost-gateway#56 – existing tests pass and does the trick, but I feel we could simplify things somehow.

@hacdias lmk if you have any idea how to make this better (if not, we can merge as-is, to unblock Retry-After, and refine later)

[hacdias]

@lidel I think the way you rewrote it is a good idea. I'm not sure how we could simplify this now, but maybe later we will figure something out. Merged for now.