relates to (part of) #42170
alternative for #42517
closes #42517

This alternative to #42517 preserves the LCOW parsing code for mounts (in case we will be using LCOW using containerd's LCOW implementation). By default, a native parser is now used (matching the host (and thus container's) platform.

After taking care of the above, tests have also be split to be per-driver, and some refactoring to improve the tests (being able to run them "parallel").

details below:

volume/mounts: move some code to correct location, and minor linting/formatting

• Remove the windowsparser.HasResource() override, as it was the same on both
  Windows and Linux
• Move the rxLCOWDestination to the lcowParser code
• Move the rwModes variable to a generic (non-platform-specific) file, as it's
  used both for the windowsParser and the linuxParser
• Some minor formatting and linting changes

volume/mounts: remove "containerOS" argument from NewParser (LCOW code)

This changes mounts.NewParser() to create a parser for the current operatingsystem,
instead of one specific to a (possibly non-matching, in case of LCOW) OS.

With the OS-specific handling being removed, the "OS" parameter is also removed
from daemon.verifyContainerSettings(), and various other container-related
functions.

Split tests per parser

Have tests per-parser, instead of grouping all tests in a single test (with a
test-set for each parser inside that test). All tests have been kept, at the
cost of some duplication in the test-cases that overlapped / were the same
for multiple parsers.

Add constructors for each parser

This adds constructors for the Linux, Windows, and LCOW, to allow using these
parsers externally, instead of only through NewParser().

Don't use global variable for fileinfoprovider

The code used a global / package-level currentFileInfoProvider that was used
to stub the actual fileInfoProvider during tests. Using a package-level variable
for this could cause issues when trying to run tests in parallel, as well as
making the tests more complicated (tests had to reset the provider in a defer).

Now that we have a constructor for each driver, we can set up the fileinfoprovider
in those constructors, and stub the implementation for specific instances, instead
of globally.

Use sub-tests, and use gotest.tools

These tests were running many test-cases, which required a lot of glue code to
construct useful errors if "one of those" failed.

Using regular sub-tests instead to make debugging easier, and use gotest.tools
to take care of useful errors.

- Description for the changelog

- A picture of a cute animal (not mandatory but encouraged)