[kernel] Rename "in_int_handler" to "blocking_disallowed"

The name "in_int_handler" is misleading because the field is only set
to true for part of the interrupt handler -- the part during which
blocking operations are disallowed.

Rename the field to reflect its actual meaning.

Bug: none
Test: build

Change-Id: Iab41d3650ea805b8de0b13de76b64c5819f48b2f

[kernel] Lockdep instrument the vm and thread_lock.

- Instrument the locks used by the vm.
- Instrument the thread lock and remainning core kernel code.
- Leaving remaining arch/lib/driver code for a later CL.

Bug: ZX-23
Test: k ut all and system/test/* pass + manual testing.

Change-Id: Ie3f6a27e81bcd1890d63f2fafda3ee233f7969ef

[kernel][thread] Add "resched_disable" count for deferring thread rescheduling

This is similar to "preempt_disable", except that it does not block
pre-emption by interrupt handlers.

We now have a single field in thread_t containing two disable counts
so that it can be checked with a single memory access.

The logic wrapping interrupt handling becomes a bit more complex. I
have moved it into functions defined in a new header, interrupt.h, so
that it can be shared across architectures and tested by unit tests.
We now save and restore preempt_pending so that we can tell whether an
interrupt handler has set it to true.

Unit tests: For each test of preempt_disable that runs outside of
interrupts, I have added a test of resched_disable.

ZX-1690

Change-Id: I4d33d5c604709a31f3c9f2d82a18e5d24ad8b5f2

[kernel] Move unittest.h to include/lib/unittest/

Move unittest.h deeper in the tree so include statements will say:

Change-Id: I85cb027a53530cd2003d813f004aaea3509f9f40

[kernel][tests] Use fbl::atomic<> rather than 'volatile'

Atomic accesses have cleaner semantics than volatile accesses, such as
including memory barriers by default.

ZX-1690

Change-Id: I5e3f47e5eaaa85fabf95a1b7cf9e7d4f42da3916

[kernel][tests] Use thread_sleep_relative() rather than thread_sleep()

Since thread_sleep_relative() is available, we might as well use it,
for readability.

ZX-1690

Change-Id: Iaf86f934eab2a477980efaa65e6c92317f480194

[kernel][tests] Avoid a possibility of flaky failure in a preempt_disable test

Remove an unnecessary assertion.

100usec is often enough that timer_set() returns before the timer has
run, but it's not always enough, in which case the EXPECT_EQ() would
fail.

Rather than increasing the wait time that we pass to timer_set(), we
can just remove this assertion, which I had originally added just to
make sure that the "while (!timer_ran)" loop works properly.

ZX-1690
ZX-1820

Change-Id: I133e1996a519e8f74075043c73ddb012a254fdc8

[kernel][thread] Allow thread_preempt_disable() outside of interrupt handlers

Change the preempt_disable field to be a counter instead of a bool,
and check preempt_pending only when preempt_disable is decremented to
zero.

ZX-1690

Change-Id: If1adefe53a23e7fb3cb4bae3a3421bb08fcb7af2

[kernel][unittest] Remove unused "void* context" argument from test cases

This argument is always NULL.

This argument was part of an init/cleanup feature that was unused and
that was removed.

Change-Id: I997a268be10c6837b2d7529fb92c2df2e3bab814

[kernel][unittest] Remove unused init/cleanup function parameters

None of the in-kernel test cases provide init or cleanup functions.
They all pass NULL to UNITTEST_END_TESTCASE() for those parameters, so
we can remove those parameters. Note that unittest.c wasn't actually
using the cleanup function.

Change-Id: I0332bc5c870deec8a021f543ec2c31ef95c30d0f

[kernel][unittest] Rename REQUIRE_* to ASSERT_* to match userland's unittest.h

This is also more consistent with gtest.

Change-Id: Ic39e75c41a55ca5cba2d09cb920c7cd1c1c7a899

[kernel][tests] Add a basic unit test for preempt_disable/preempt_pending

This just tests that preempt_disable is set for timer callbacks, and
that preempt_pending will get set by some functions in this context.

ZX-1690

Change-Id: I41c6d698272a85abe27732121e0fdb0096b2a3dd