Asset Browser from Python API point of view

This document is not exhaustive, and just serves as a scratch pad where I can jot down things.

Missing Python API functionality

This is what I (Sybren) found missing, but what I'd like to have available for making the Pose Library prototype:

Getting active asset in Asset Browser (currently only works for 'current file' assets, i.e. when the asset is an already-loaded datablock), i.e. enough information to:
- know what the type of the asset is (so currently the ID type), and
- to load the datablock itself.
Getting information from all available assets (for example to get all tags).
A way to refresh the Asset Browser (for example after creating a new asset).
Setting active asset in Asset Browser (for example after creating a new asset).
Getting list of selected assets (for example for mass-rename or mass-tagging).
Callback when active asset changes. This is necessary to enable "follow mode".
Callback when asset is double-clicked.

Action datablock with pose must be loaded into memory, in order to apply the pose. There are multiple ways in which a datablock can be loaded:

Linking:
- Pros:
  - Most natural approach.
  - Keeps link with original asset file, so reloading is possible.
  - Reloading operator already available.
- Cons:
  - Pollutes the blend file namespace.
  - Link will be saved with the blend file. This is permanent (until cleaned up) and will produce errors when the original pose library is no longer available.
  - Needs reloading of links when pose library changes.
Appending:
- Pros:
  - No link to original asset file, so will not cause errors when asset file is no longer available.
- Cons:
  - Pollutes the blend file namespace and expands the blend file.
  - Unable to reload.

Either approach can be used in a temporary fashion, like such:

The last two steps can be avoided by using a non-main datablock (i.e. a datablock that will not be part of the blend file itself) for the asset.