Add storage helper and migrate config entries

[balloob]

Description:

This adds a storage helper. This will make it easier for components and core pieces to store and load data.

• Stored data needs to be JSON serializable. However when and where the config is stored is opaque to the developer. This leaves options open for the future 👍
• Optionally add a write delay in case you expect a lot of writes
• If delayed write is scheduled, load will return correct config
• If delayed write, shutting down Home Assistant will trigger the write
• Add a migrator from old formats so that it helps existing code migrate to the storage helper.
• Track versioned data and migrate if necessary

Since I did not want to add logic that is unused, I've migrated the config entries logic to use the new storage helper.

Bonus:

• Fixes Add async_add_executor_job function to improve performance and avoid cpython bug  architecture#37 (comment). Adds a new hass.async_add_executor_job if we know it's going to be a sync method. This will be the new preferred way for sync methods. Will help step towards Make typing checks more strict #14429 (comment)

Checklist:

• The code change is tested and works locally.
• Local tests pass with tox. Your PR cannot be merged unless tests pass

If the code does not interact with devices:

• Tests have been added to verify that the new code works.

[balloob]

To do: pass in version and migrate function to store constructor -> DONE

[pvizeli]

we should take that toggether like:
except (json.xy, json.ser) as err:

[pvizeli]

you use for all async function async. That is the only function they use a other name concept

[pvizeli]

Future: It would be nice to add a voluptuous schema for data partition. That make it easy to add defaults or changes over the schema.

[balloob]

I don't want to include voluptuous, the idea is that this is written/read by code, not humans. If defaults change, they can do a one time migration to create those. It should not run on every run.

[pvizeli]

For simplify the code, I suggest to register the STOP event once and if they is reach, we can check if data are available and write it down or do nothing. So we need only handle the delay listener and can cleanup a lot of logic and code.

[balloob]

I don't want to register the STOP event during initialization because it means that startup of Home Assistant is slower. With the current logic, although it looks a bit complicated, we make sure that we only register handlers when we need it and and as little as possible.