tests: Ensure failing test with random fixture data can be reproduced

Some tests input data returned by swh-web fixtures are sampled from
filtered objects stored in a static test archive.

In order to reproduce any failing test with the same fixture data:

• add a new autouse function scope fixture setting the random seed to current time before each test execution and putting that seed in cache

• implement custom pytest reporting which adds a new section detailing the random seeds used for each failing test and the pytests commands to reproduce the failures

• add a new pytest option --swh-web-random-seed enabling to explicitly specify the random seed that will be used to run the tests