Add critical tests to the workflow and block failed PR

[choisant]

As a first step towards making a proper suite of tests, we should make some basic tests. These tests should check the main functionality of the software and the main results of the calculations. They should be automatically run when a pull request is opened, and the pull request should not be allowed to be merged if the tests do not pass. Check when each is done, then label this issue as completed. Edit the suggestions as needed. Non-critical tests should be added to the Issue related to those.

• Does the installation work?
• Does parallel processing work where the test is run?
• Can each of the main functions run without error?
• For a specific data sample, does the buildmetadata() function work as expected?
• For a specific data sample, metadata file and random seed, does the mcmcoutput stay consistent within a margin of error?
• For a specific learnt output and test data sample, do the mean probability and quantiles stay consistent when calculated over the whole output?