This includes Carbon support

This is intended to fix two issues:
- In REAPER, setting automation recording to 'Latch', recording some
  automation, saving and reloading the project, then starting playback causes
  automation data to be overwritten. This appears to be because REAPER
  interprets parameter change messages sent during state restoration as
  originating from the user.
- In Ableton Live, automation lanes are sometimes disabled when loading
  projects. This also seems to be because the setStateInformation call
  may send parameter change messages back to the host.

According to the docs for kAudioUnitProperty_ClassInfo, the host is
required to notify all parameter listeners that the parameter values may
have changed, implying that the plugin need not send its own parameter
change notifications during state restoration.

Some AUv3 presets crash when querying the set of presets in Loopy Pro.
The issue seems to be because `addPresets` may end up being called
concurrently with the host's queries.

Some AUv3 presets crash when querying the set of presets in Loopy Pro.
The issue seems to be because `addPresets` may end up being called
concurrently with the host's queries.

Fixes an issue introduced in 72fa2d98e1
where zeromem might write past the end of the channel buffer.

This autorelease was accidentally removed in
2a199d5225

Note that hosts using the AUv2 API (e.g. JUCE hosts) to host JUCE AUv3
plugins may not receive begin/end gesture events, depending on the OS
version, and potentially the host architecture. I suspect this is
because older versions of the OS-provided AUv2/AUv3 translation layer
don't handle these events.

In testing, an Intel host on Catalina did not receive begin/end events,
but an Arm host on Monterey did receive these events.

Previously, with two instances of the MIDILogger AUv3 in series in
Logic, the timestamps received by the second plugin in the chain would
not match the timestamps of the events emitted by the first plugin.

This change allows mouse events (including enter and exit events) to
reach unfocused views on macOS. This matches the behaviour of unfocused
windows on Linux and Windows, where components paint in their "hovered"
states even when the application window is in the background.

As a byproduct of using tracking areas on macOS, we can remove the fake
mouse move generator.

Ideally, we want to pass shortcut keys to the component to handle, and
only fall back to invoking a menu item if the component was unable to
handle the keyboard event, or if the action was triggered by
clicking/selecting an item in the menu itself. The old implementation
tried to work out whether the action was triggered by a shortcut by
checking the event's characters and modifiers. This method was
inaccurate, because some shortcuts (such as arrow keys) may add
unexpected numpad/function modifier flags.

We now try handling shortcut keys directly in the peer, and pass events
up to the superclass (which will forward them to the main menu) if the
event could not be handled.

This commit also adjusts some Objective-C method signatures to use the
correct string encoding for the BOOL type.

MIDI FX AudioUnits can retrieve their current sample rate from an output
bus. However, it's reasonable for a JUCE MIDI FX plugin to require 0
output buses, in which case the correct sample rate cannot be retrieved.

This patch modifies the AU wrapper to ensure that MIDI FX plugins will
always have at least one output bus. This bus does not necessarily
correspond to a matching bus in the AudioProcessor, and it will not be
passed to AudioProcessor::processBlock. With the output bus in place,
the correct sample rate is reported in `prepareToPlay`.

This patch also moves PluginHostType into juce_audio_processors in order
to allow hiding of function implementations.

Reports the correct host name for Apple hosts on M1 which host plugins
in a separate process.

Affected hosts are:
- Logic Pro
- GarageBand
- MainStage
- Final Cut Pro