feat: provide interface frida.Device

[dylanhitt]

Provide the interface of frida.Device. Pretty similar to what was just done with frida.DeviceManager electing to only return the interface to the caller.

Discussion point:

You may notice I elected to change the implementation over to DeviceImpl. I did this so it would preserve the current godoc. In the case of DeviceManager I forgot that unexporting implementation would not allow for godoc parsing. For example you will not find any of the functions on the latest godoc: https://pkg.go.dev/github.com/frida/frida-go@v0.7.0/frida#DeviceManager. You will only find the interface.

I would say this leaves two options.

1. Reexport deviceManager as DeviceManagerImpl (like how this MR does it).
2. Take the approach the stdlib does with crypto/cipher: https://pkg.go.dev/crypto/cipher#AEAD. Where the docs are provided in the interface definition. (If this is the option chosen. I'll change this PR to reflect and unexport DeviceImpl as device).

I'm not married too either direction. Number 2 doesn't feel too great from the end user side. Number 1 does expose the struct, but in reality most users would just call the New func and go about their business. However, either I think adding comments with the interface (like number 2) is a nice touch in general and would be nice to do.

[dylanhitt]

Changed this up a bit after looking through other CGO libs/the std lib/reading. I still think adhering to the go.dev suggestion that is important:

Do not define interfaces on the implementor side of an API “for mocking”; instead, design the API so that it can be tested using the public API of the real implementation.

Which I don't mind slapping an interface into packages, IMO the caller should be able to do whatever they want. Whether they want to use the concrete type and declare their own interface or use the one provided by the package.

However do to the nature of this API returning an interface in the case of device from DeviceManager/Portal feels really nice (as there is literally no other way for the caller to mock the functionality easily. Returning the concrete type in this case requires the user to wrap everything returned with their own struct which doesn't feel good at all). I think this is where the go proverb of "CGO is not go" comes into play 😆.

Looking forward to the feedback. Apologies for the spam from working through this.

[NSEcho]

Hi and thanks for such big changes. I like this one more than the previous one honestly, because it kinda keeps the same semantic without breaking any functionality.

[dylanhitt]

Yes, I hate having to suffix interfaces with Int. But whatcha gonna do. Cheers!