Even if snapshots are supported by the driver interface, the driver or
suite might still want to skip a particular test, so those checks
still need to be executed.

This addresses the two remaining change requests from
kubernetes#69036:
- replace "csi-hostpath-v0" name check with capability
  check (cleaner that way)
- add feature tag to "should create snapshot with defaults" because
  that is an alpha feature

Signed-off-by: Patrick Ohly <[email protected]>

There is no need to check for empty strings, we can also directly
initialize structs with the value. The end result is the same when the
value is empty (empty string in the struct).

When the provisioning test gets stuck, the log fills up with messages
about waiting for a certain pod to run. Now the pod names are
pvc-[volume-tester|snapshot]-[writer|reader] plus the random
number appended by Kubernetes. This makes it easier to see where the
test is stuck.

TestDynamicProvisioning had multiple ways of choosing additional
checks:
- the PvCheck callback
- the builtin write/read check controlled by a boolean
- the snapshot testing

Complicating matters further, that builtin write/read test had been
more customizable with new fields `NodeSelector` and
`ExpectUnschedulable` which were only set by one particular test (see
kubernetes#70941).

That is confusing and will only get more confusing when adding more
checks in the future. Therefore the write/read check is now a separate
function that must be enabled explicitly by tests that want to run it.
The snapshot checking is also defined only for the snapshot test.

The test that expects unschedulable pods now also checks for that
particular situation itself. Instead of testing it with two pods (the
behavior from the write/read check) that both fail to start, only a
single unschedulable pod is created.

Because node name, node selector and the `ExpectUnschedulable` were
only used for checking, it is possible to simplify `StorageClassTest`
by removing all of these fields.

Expect(err).NotTo(HaveOccurred()) is an anti-pattern in Ginkgo testing
because a test failure doesn't explain what failed (see
kubernetes#34059). We avoid it
now by making the check function itself responsible for checking
errors and including more information in those checks.

Whether the read test after writing was done on the same node was
random for drivers that weren't locked onto a single node. Now it is
deterministic: it always happens on the same node.

The case with reading on another node is covered separately for test
configurations that support it (not locked onto a single node, more
than one node in the test cluster).

As before, the TestConfig.ClientNodeSelector is ignored by the
provisioning testsuite.

This is a special case that both kubelet and the volume driver should
support, because users might expect it. One Kubernetes mechanism to
deploy pods like this is via pod affinity.

However, strictly speaking the CSI spec does not allow this usage
mode (see container-storage-interface/spec#150) and
there is an on-going debate to enable it (see
container-storage-interface/spec#178). Therefore
this test gets skipped unless explicitly enabled for a driver.

CSI drivers which create a block device for a remote volume in
NodePublishVolume fail this test. They have to make the volume
available in NodeStageVolume and then in NodePublishVolume merely do a
bind mount (as for example in
https://github.com/kubernetes-sigs/gcp-compute-persistent-disk-csi-driver/blob/master/pkg/gce-pd-csi-driver/node.go#L150).

The driver should support multiple pods using the same volume on the
same node.