Import the whole string library, as well as the ctype functions from the musl
C library.  Must is a BSD licensed library that aims to be API and mostly ABI
compatible with glibc, so it's our easiest way to provide the ABI we need
without reinventing the wheel.  I will import more of the traditional non-syscall
C library code from it and expect libc/ to eventually mostly consist of slightly
modified musl code.

After my recent test suite changes tests are expected in /tests

Lots of work towards enabling pthread_cond_timedwait()

Conflicts:
	bootfs.manifest
	build.mak
	loader.cc

thread::wake() may be called from an interrupt handler (clockevents).

The thread that's doing the waking is obviously running, so no need to
reschedule, and it's awkward if done from an interrupt.

Thread timers can be set to wake a thread at a specified time, or cancelled.

Not yet preemption safe.

Fox simplicity, the first mutex implementation will not be compatible with
PTHREAD_MUTEX_INITIALIZER, so wrap the mutex class with a structure that also
contains an intitialization flag, and initialize it on demand.

Missed adding the object base.

It's actually percpu, but we don't have percpu support yet.  So make it
global.

Way too noisy

Allow the clockevents driver to use it during startup.

This is intentionally simplistic, supporting only one callback (intended
for waking up threads).

Use:

   with_lock(irq_lock, [] {
      stuff done with interrupts disabled, including early return stmts
   });

There now is a public add_object and a private set_object method, which
describe the two totally different actions much better.

Style requires manual edititing of loader.cc to run, no automated way
to pick up whole directories in the manifest, and we won't clear the
environment before running the next .so

Added pci config capability parsing and moved the virtual call to earlyInitChecks() to the base class

found by the glibc testsuite

Fix various corner cases to pass the glibc test suite.

This makes the apic immediately usable.

mmap() did not consider that a large page may be used to map an area that
it is overwriting.  This can happen if the primordial 1:1 area is remapped
(as when the payload is creating some guard pages).  The result is that
random memory is overwritten with a pte.

Fix by detecting large pages and splitting them up.