The Great Crate Re-organization

[Carter12s]

Why?

1. The purpose / usage of the crate has morphed over time. Original use case was to talk to rosbridge. Current active use case is to provide generic way to talk over multiple protocols.
2. We're still "leaking dependencies" in strange ways.
   a. Want to use roslibrust_codegen_macro, well the generated code assumes you also depend on roslibrust_codegen
   b. Want to just use roslibrust_zenoh? Gotta depend on traits defined in roslibrust...
3. Common usecases require compiling a ton of stuff you don't use. If you're not using the rosbridge backend why are you compiling it.

The new plan:

Ramifcations / API changes

• Users will need to update feature flags. I'm thinking the rosbridge backend won't be included in the default set
• Rosbridge backend will move out of the global namespace of the roslibrust crate. So it will be roslibrust::rosbridge::ClientHandle instead of roslibrust::ClientHandle`
• Many places where users are currently needing to depend on roslibrust_codegen they'll be able to remove it

Eventually I'd like to provide roslibrust_mqtt and roslibrust_nats...

[Carter12s]

Starting feeling this out in this MR: #215

[Carter12s]

As I've worked on this I realize it is slightly more complicated to achieve the exact presentation I want.

What I'd like is that users can directly depend on "roslibrust" with feature flags to enable specific backends (mock, ros1, rosbridge, zenoh).

To prevent a circular dependency between roslibrust and the backend crates, the Traits for TopicProvider have to be declared in common ancestor crate.

Updated plan is: