Introduce gevent request timeout

[varmar05]

Currently we do not have any bounds in case we run server with gevent worker type. In case client gave up on connection server still might be processing request. Also if request is greenlet safe (as it should be!) it can take long time to finish since it yields to gevent hub and standard gunicorn timeout is not applied here.

This PR introduces custom timeout via middleware to cut the request early enough in case of gevent. There are two new related variables GEVENT_WORKER and GEVENT_REQUEST_TIMEOUT.

Logs from a dummy test

[ERROR] [23194] Gevent worker: Request /ping timed out
[ACCESS] - GET /ping  HTTP/1.1 502 11 - "Mozilla/5.0 (X11; Linux x86_64) ..." 1003113 <23194>

[coveralls]

Pull Request Test Coverage Report for Build 11819057268

Details

• 66 of 71 (92.96%) changed or added relevant lines in 35 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.05%) to 93.765%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
server/mergin/tests/test_middleware.py	16	17	94.12%
server/mergin/app.py	15	19	78.95%

Totals
Change from base Build 11816539068:	-0.05%
Covered Lines:	6376
Relevant Lines:	6800

---

💛 - Coveralls

[mostlyAtNight]

a few comments to go through but looks great