Add tests for object releasing

[almarklein]

Closes #353

This adds tests, in a new folder so it won't interfere with the unit tests. The idea is basically, doing for each kind of object:

• Get how many objects we have of each kind.
• Create a bunch of objects of a specific kind.
• Make sure the counts are indeed up.
• Delete the objects.
• Check that the counts are down.

Todo:

• Generic logic that is the same for each unit test.
• Implement some meta-tests to make sure that the tests indeed fail when they should.
• Implement a test function for every object.
• Can we now remove that delayed_releaser? Yes!
• Can we now enable some release calls that previously panicked? Yes!
• Measure process memory consumption to detect leaks? --> Test against memory leaks by measuring process memory #406
• Do we want our BindGroupLayout object to cache instances like wgpu-core does?
• Does this close Rust panic after creating hundreds of renderers/canvases #275?
• Investigate fails:
  • Device (simply does not clean up in native).
  • Queue (related to the above)
  • No luck yet.
  • Following up in Device objects are not cleared on wgpu-native #405

[Korijn]

Can you explain what polling the device does conceptually?

[almarklein]

Can you explain what polling the device does conceptually?

From what I understand, wgpu has an an event-loop of sorts, so it can perform asynchronous tasks. Polling the device flushes the device's queue. We also use poll in the functions that are asyn in nature, like buffer mapping. Related to #391.

[almarklein]

I think this is good to go. I'll have a look (probably after the weekend) what the state of #275 is. But in the mean time this is ready for review.

[almarklein]

I created 2 new issues for follow-up work:

• Device objects are not cleared on wgpu-native #405
• Test against memory leaks by measuring process memory #406