Improve record/replay tests

[Porges]

A few changes here:

Check bodies in requests

This makes sure the request/response pair is for the correct request (by default, go-vcr only checks the HTTP method/URL). To achieve this we blank out any dates/timestamps that are found in bodies, since they cause a mismatch when the test is rerun.

Fix "count" header addition

We had code to track the request count at a per-URL level so that we handle multiple requests to the same URL properly in recordings (so that a series of polling GETs can get back different responses, for example), but this was not being added properly. Now it is.

Better messages when go-vcr interaction cannot be found

At the moment when running record/replay tests and an interaction (request+response) cannot be located, it is hard to figure out what went wrong. This PR adds a wrapper around the go-vcr code to add more context about what went wrong.

An example message in the output looks like:

*** Cannot find go-vcr recording for request (no responses recorded for this method/URL): DELETE https://management.azure.com/subscriptions/00000000-0000-0000-0000-000000000000/resourceGroups/k8sinfratest-rg-kjqcwl?api-version=2020-06-01

Ideally we would panic when this happens, but we can't at the moment (the code explains why).

Closes #1432.

How does this PR make you feel:

[codecov-commenter]

Codecov Report

Merging #1492 (40661b9) into master (b790d84) will increase coverage by 0.07%.
The diff coverage is 59.64%.

@@            Coverage Diff             @@
##           master    #1492      +/-   ##
==========================================
+ Coverage   62.13%   62.20%   +0.07%     
==========================================
  Files         161      163       +2     
  Lines       10695    10730      +35     
==========================================
+ Hits         6645     6675      +30     
- Misses       3414     3415       +1     
- Partials      636      640       +4     

Impacted Files	Coverage Δ
...erated/pkg/testcommon/kube_test_context_envtest.go	84.00% <ø> (+8.09%)	⬆️
hack/generated/pkg/testcommon/test_context.go	58.82% <52.94%> (+3.94%)	⬆️
...d/pkg/testcommon/error_translating_roundtripper.go	53.12% <53.12%> (ø)
.../generated/pkg/testcommon/counting_roundtripper.go	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b790d84...40661b9. Read the comment docs.

[matthchr]

LGTM - left a few comments/questions though.