Player API: Animation-dependent collisionbox and eye_height

[paramat]

Closes #2742
See #2742 (comment) and following comments.
For the registered MTG player model, collisionbox and eye-height will now be correctly set when 'sitting' or 'laying'.
The main benefit however is for mod-registered player models and their various and possibly more unusual animations, for example a crawling animation could now allow a player to pass through 1-node high spaces.

Player model collisionbox and eye_height can each be defined as a table of per-animation values, or left as a single universal value (which also ensures backwards compatibility).
When animation is set in set_animation(), the animation-dependent object properties are also set.

The set_animation() function returns if the requested animation is already applied to a player, so the function only executes when the animation needs to change, for example starting walking, stopping walking etc.

Now that object properties are also set in set_animation(), an object properties packet is sent to the client each time animation changes. I hope this is not a significant increase in network traffic, as it will be on average 1 extra packet every second (very approximately).

[paramat]

Testing ...

^ Dead player collision/selection box

^ The new view when dead, close to the ground

^ Sitting player collision/selection box

^ The new view when sitting

[appgurueu]

This PR is fine content-wise, but IMO player_api needs a redo: table keys should not be duplicated - instead, subtables should be used. Example:

animations[animname] = {x = ..., y = ...}, collisionbox[animname] = {...}, eye_height[animname] = {...}

would become

animations[animname] = {x = ..., y = ..., collisionbox = {...}, eye_height = ...}

[paramat]

Thanks for the input.

table keys should not be duplicated

I accept that might be better coding practice, but it is not important or essential.
Your suggestion requires radical changes to player_api. Familiarity, simplicity, minimal change and minimisation of compatibility risk is perhaps more important.

[paramat]

PR is ready.
Some collisionbox and eye height values still need fine-tuning, but that will be done very soon and before a merge, the PR is otherwise complete, reviewable and testable.

[paramat]

Added a second commit that avoids setting properties if there is no change in collisionbox, which is most of the time for the MTG player model, as the collisionboxes for stand, walk, mine, walk_mine are all identical.
A change of eye_height is not checked, it is assumed that eye height does not change if collisionbox does not change, this is the case for the MTG player model and will almost always be the case for all models. This assumption simplifies the code.
Retested.

[paramat]

Closing, #2745 is better.