libct/cg: refactor/improve/rename GetHugePageSize -> HugePageSizes

[kolyshkin]

(this is based on and currently includes PR #3233)

1. Since GetHugePageSize do not have any external users (checked by
   sourcegraph), and no internal user ever uses its second return value
   (the error), let's drop it.

2. Rename GetHugePageSize -> HugePageSizes (drop the Get prefix as per
   Go guidelines, add suffix since we return many sizes).

3. Implement lazy init for HugePageSizes

   I have noticed that libct/cg/fs allocates 8K during init on every runc
   execution:

   init github.com/opencontainers/runc/libcontainer/cgroups/fs @1.5 ms, 0.028 ms clock, 8512 bytes, 13 allocs

   Apparently this is caused by the global HugePageSizes variable initialization,
   which is only used from GetStats (i.e. it is never used by the runc binary itself).

   Remove it, and use HugePageSizes() directly instead. Make it init-once,
   so that GetStats (which, I guess, is periodically called by kubernetes)
   does not re-read huge page sizes over and over.

   This also removes 12 allocs and 8K from libct/cg/fs init section:

   $ time GODEBUG=inittrace=1 ./runc --help 2>&1 | grep cgroups/fs
   init github.com/opencontainers/runc/libcontainer/cgroups/fs @1.5 ms, 0.003 ms clock, 16 bytes, 1 allocs

4. libct/cg: TestGetHugePageSizeImpl: use t.Run

   Move test case comments to doc strings, and use t.Run.

[kolyshkin]

This is definitely post-1.1 material

[kolyshkin]

$ time GODEBUG=inittrace=1 ./runc --help 2>&1 | grep cgroups/fs

Speaking of this, I have a WIP set of patches to remove some init-time allocations.

[thaJeztah]

LGTM

left one suggestion (could be in a follow-up)

[thaJeztah]

Looks like the only place the error returned from getHugePageSizeFromFilenames() is used, is in a unit test. That test already checks for warnings; wondering if we should change the errors to warnings (as we won't be handling them);

runc/libcontainer/cgroups/utils_test.go

Lines 534 to 537 in 5fb9b2a

	// check for warnings
	if c.isWarn && warns.Len() == 0 {
	t.Errorf("input %v, expected a warning, got none", c.input)
	}

While at it, wondering if we should change that test to use the logrus testhook, like

runc/libcontainer/capabilities/capabilities_linux_test.go

Lines 24 to 25 in 0d49470

	hook := test.NewGlobal()
	defer hook.Reset()

[kolyshkin]

OTOH we can simplify the whole thing, letting getHugePageSizeFromFilenames() always return errors, and log them as warnings in HugePageSizes()!

....

Implemented, PTAL @thaJeztah

[kolyshkin]

Rebased on top of current git tip to repro #3266

[thaJeztah]

code LGTM

is the failure still the thing we're debugging?

[kolyshkin]

is the failure still the thing we're debugging?

No, it was a one time thing, totally unrelated, and fixed by #3273

[kolyshkin]

Since 1.1 is delayed I think this can be included.

[thaJeztah]

LGTM!

[kolyshkin]

@AkihiroSuda PTAL