feat: new timeout writer implementation

[hf]

#1529 introduced timeout middleware, but it appears from working in the wild it has some race conditions that are not particularly helpful.

This PR rewrites the implementation to get rid of race conditions, at the expense of slightly higher RAM usage. It follows the implementation of http.TimeoutHandler closely.

[coveralls]

Pull Request Test Coverage Report for Build 9161432738

Details

• 60 of 75 (80.0%) changed or added relevant lines in 3 files are covered.
• 3 unchanged lines in 2 files lost coverage.
• Overall coverage decreased (-0.04%) to 65.7%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
internal/api/verify.go	10	11	90.91%
internal/api/middleware.go	49	63	77.78%

Files with Coverage Reduction	New Missed Lines	%
internal/api/verify.go	1	72.55%
internal/api/middleware.go	2	79.85%

Totals
Change from base Build 9124537519:	-0.04%
Covered Lines:	8401
Relevant Lines:	12787

---

💛 - Coveralls

[kangmingtay]

@hf i've updated the branch with the following changes:

1. Verify email change shouldn't be calling http.Redirect twice
2. timeoutMiddleware doesn't have to be a struct method of API since we don't use any of the properties in it
3. Calling WriteHeader now takes a snapshot of the existing header map. When finallyWrite is called, we'll write the header from the snapshot rather than the header map.

[J0]

Thanks!