Move client functionality into client

[ibeckermayer]

There are no functional changes here, this is just renaming things (all FFI functions which accept a *const Client are prefixed with client_ and moving core Client logic into Client methods, rather than pushing that logic out to the FFI functions themselves.

The FFI client_ functions now have sole responsibility of converting the *const Client to a &Client and calling the appropriate method. This prepares me to move Client into its own module.

[espadolini]

Since C has no namespacing I would suggest using rdpclient_ as a prefix, as client_ might be a bit too generic.

[espadolini]

I'd expect this to just return a Result<Self, ConnectError> and the FFI wrapper can deal with boxing and returning it as a pointer, to move the actual implementation away from FFI details.

[ibeckermayer]

I don't agree, the Client is fundamentally an "FFI thing" that lives life as a *const Self, it's created to be passed back over the FFI and has the methods from_raw and drop which (more or less) take this "FFI thing" as a receiver

/// # Safety
    ///
    /// client_ptr MUST be a valid pointer.
    /// (validity defined by https://doc.rust-lang.org/nightly/core/primitive.pointer.html#method.as_ref-1)
    pub unsafe fn from_raw<'a>(client_ptr: *const Self) -> Result<&'a Client, CGOErrCode> {
        client_ptr.as_ref().ok_or_else(|| {
            error!("Client pointer is null");
            CGOErrCode::ErrCodeClientPtr
        })
    }

    /// # Safety
    ///
    /// Calling this twice on the same Client
    /// pointer will result in a double free.
    pub unsafe fn drop(ptr: *mut Self) {
        if !ptr.is_null() {
            drop(Box::from_raw(ptr))
        }
    }

That said, it looks like we may be able to get rid of this *const grossness at some point, see #29764 (comment)