meson.build:
When the 'wrap_mode' option is set to 'nodownload' use a system version
of cpptoml. This does not require using git and having a network
connection during build, which is important for Linux packaging
infrastructure.

subprojects/cpptoml.wrap:
Pin revision to last release tag (v0.1.1).

lib/wptoml/*.cpp:
Remove 'include/' prefix from all cpptoml related includes, at is not
required.

Closes #17

1. device export proxies must be destroyed manually since they are
   not associated with the WpRegistry
2. the monitors should not disconnect before all WpSpaDevice objects
   are destroyed; remove the manual disconnect call and let GObject
   ref counting do its job (the core will disconnect when its last ref
   count is dropped after the last monitor plugin is destroyed)

otherwise the parent never un-exports or de-activates itself
and we leak a bunch of stuff

it requires GLib 2.60 and it's not worth it

g_object_get_property did not support uninitialized GValues prior
to GLib 2.60, and it's not worth requiring 2.60 for this reason

Fixes #11

- pw_core_info has data that represent the remote core
- pw_core properties are the properties of its pw_client,
initially inherited from the pw_context, which can be updated either
when calling pw_context_connect() or later at runtime (the pw_core
calls pw_client_update_properties() on its pw_client)

wp_core_update_properties() is made in such a way so that we can do:
```
clone = wp_core_clone(core);
wp_core_update_properties(clone, ...);
wp_core_connect(clone);
```
and get clone to have different properties than the original core,
while they still share the same pw_context underneath

also delay the "connected" signal until info is there

a method to take the internal pw_properties without copying

and do not allow writing to wrapped pw_properties,
since they are const now

the spa object undereath doesn't do that and we lose a bunch of
properties otherwise

By mistake, WpImplNode was developed by keeping in mind that the proxy
returned by pw_core_export() is a PW_TYPE_INTERFACE_Node, but this
is not true. It's actually a ClientNode...

Unfortunately, making WpImplNode work as if it was a WpNode is
not so easy, especially when it comes to handling params, which
need to be queried syncrhonously on the underlying spa_node.

So, instead of fixing WpImplNode to work as a WpNode, we choose to
disconnect them. This way, WpImplNode will not be used as a proxy
in the registry and the registry will normally create WpNode proxies
instead, making round-trips through the server to change node params.

This is necessary to finish augmenting nodes that advertise
PropInfo & Props in their param info but they don't actually have
any properties, so there is no param event emitted
(ex. the jack device nodes)

Because the sync is synchronous with the calls over the protocol,
the callback is ensured to be called after all the param events
have been emitted

otherwise we get stuck waiting for the param event
and we get warnings on the debug log

this allows the created nodes to be able to set their "device.id"
property correctly

they are not useful anymore because we hijack priv->info from the
impl subclass, so that it points to the impl info struct,
and therefore the base implementations work just fine

This reverts commit 917a019b.