Add a FlatShape interface to allow manipulating the detected shapes in a unified way

[juliabeliaeva]

Adding a common interface allows to simplify writing code like this:

private fun <T : FlatShape<T>> T.flip(): T {
    return map { x, y -> 1 - x to y }
}

private fun <T : FlatShape<T>> T.flipIfNeeded(isImageFlipped: Boolean): T {
    if (isImageFlipped) return flip()
    return this
}

Flipping detection results horizontally is used on Android when the front camera is used.

[juliabeliaeva]

At the same time, this could be done with the when statement, and when is still used when drawing detection results in the Android demo app. So I'm not sure how necessary this is. @ermolenkodev what do you think?

[ermolenkodev]

I like it! I think the nice API is the main selling point of this library so this change fits perfectly