[tracing] Add more logging of errors

Plus add some clarifying comments to provider.h.

Tested: runtests
Plus observe some new error messages while debugging DX-448

Change-Id: Iba682eeab128326cf32e5e9c0f90ba774356b1b4

[trace-provider] Add RegisterTraceProviderSynchronously

... and the ability to wait for tracing to start.
This allows programs like benchmarks to avoid having early trace
records dropped due to the benchmark starting to run before TraceManager
has sent the Start request.

DX-164 #comment patch

Tested:
trace record /pkgfs/packages/trace_stress/0/bin/trace_stress

Change-Id: I4c8e9e4cd37587e2624879fcc23720f525349238

[zircon-internal] Move fnv1a out of zircon/public

We don't need to expose this hash function in the Zircon system
interface. We can put it in a library instead.

Test: Compiles.
Change-Id: I3a6f63997f845cf5ae78f969326d3e650950601d

[trace] Build with -fvisibility=hidden

ZX-1895 #comment

Test: No behavior change
Change-Id: Iaf82522cb8c74241b55d542d77d1d900ce8d0da4

Revert "[hash] Move fnv1a out of zircon/public"

This reverts commit 062af18fa8867ef61816072bc939ad8bff30d7d2.

Garnet build does not understand header-only deps.

Original change's description:
> [hash] Move fnv1a out of zircon/public
>
> We don't need to expose this hash function in the Zircon system
> interface. We can put it in a library instead.
>
> Test: Compiles.
> Change-Id: Ife133b7e4e4325dade11b5e3a7d1b82aa6899ece

TBR=kulakowski@google.com,swetland@google.com,jeffbrown@google.com,abarth@google.com

Change-Id: Iac2ce9a0245b4f5788e09118209657bfc6448c52
No-Presubmit: true
No-Tree-Checks: true
No-Try: true

[hash] Move fnv1a out of zircon/public

We don't need to expose this hash function in the Zircon system
interface. We can put it in a library instead.

Test: Compiles.
Change-Id: Ife133b7e4e4325dade11b5e3a7d1b82aa6899ece

[tracing] Port trace provider to fidl2.

A bit hacky for now pending improvements to the fidl2 compiler.

Change-Id: I944a22ad002f957cb2ac13fb605ba01f3fb609f7

[async] Separate fbl dependency from async_t

The C++ FIDL support library in Garnet wants to be able to use async_t
without linking against libfbl. Previously, the C declarations needed
create an async_wait_t were in the same header as code that depended on
fbl, which meant C++ clients of async_wait_t were forced to depend on
fbl.

After this change, the C++ wrappers for libasync interfaces are
separated from the core C declarations.

Change-Id: Ic172911a9dc775d5b5e540a33ce87d2dc218a3d9

[build] start exporting package descriptions for source packages

Change-Id: Idd4e916dcfee4d9f40cbb82827042b746c8ce987

[build][ulib/hash] move hash/rand functions to public headers

Instead of maintaining a library for these,

hash.h -> zircon/misc/fnv1hash.h and zircon/misc/xorshiftrand.h

Change-Id: Ia4c1a547b10f6444acd89f72c3ed340a4210f30b

[ulib][trace-provider] Hack in support for categories

... until we can use fidl-generated code.

MG-1121 #done

Change-Id: I26885fe5856522b6806dc9d33a6f6567eb70eb46

[zx] Magenta -> Zircon

The Great Renaming is here!

Change-Id: I3229bdeb2a3d0e40fb4db6fec8ca7d971fbffb94

[zircon][mxtl->fbl] Rename mxtl to fbl

Change-Id: Ie21b6498e1bfb0a7fa0315e40b9e5c3ee78646be

[trace] Port Fuchsia tracing system to Magenta.

The Fuchsia tracing system supports instrumentation of userspace
programs to collect trace records.

This implementation consists of...

- libtrace: a static library providing C and C++ instrumentation macros
and functions designed to be linked into client programs
- libtrace-engine: a shared library with a C ABI which manages
global tracing state in the process and provides
functions to write trace records
- libtrace-provider: a static library which implements the TraceProvider
FIDL interface and registers an instrumented
program with the trace manager
- libtrace-reader: a static library which provides a C++ API for reading
binary trace data
- unit tests
- benchmarks
- documentation

This is almost a complete rewrite of the code formerly in //apps/tracing
but it remains compatible with the trace format and preserves most of
the structure of the trace event API.

The main differences are...

- The <trace/event.h> API uses identical syntax for C and C++ with
a few extra goodies for C++ type inference.
- The trace engine is factored out as a shared library with a C ABI
instead of being written as a C++ static library, this makes it more
suitable for instrumentation of libraries and device drivers.
- The API is extensively documented.
- We actually have benchmarks to measure instrumentation overhead.

There are still a few pieces left in //apps/tracing, such as the trace
manager and command-line tools, which will be moved over later.

Change-Id: I2d2986be0e63e5aa21fe56dd03024490b0334214