Track active tools and SDKs in activated.json

[sbc100]

Store active tools and SDKs in a dedicated activated.json file when they are activated by emsdk activate. Update Tool.is_active() to check against activated.json rather than matching entries in .emscripten.

This solves several issues with the previous .emscripten-based is_active() logic:

1. Tools with no activated_cfg (such as ccache or tools that only modify PATH or environment variables) previously returned is_active() == False because they had no .emscripten keys and no dependencies (len(deps) == 0).
2. Tools with dependencies but no activated_cfg previously falsely reported is_active() == True whenever their dependencies were active, even if the tool itself was never activated.

This allows us to remove all of the activated_cfg entries that were there only to keep track of tool activation.

Replaces: #1345

[kripken]

The new behavior seems reasonable, but also changing the current long-standing behavior seems risky. How much of a benefit do you estimate our users would get from this?

[sbc100]

I think its much cleaner and clearer design. Without this we need to have dummy config vars in the .emscripten file that just server as magic strings to show that a tool is active. Its odd at best and confusing at worst.

For example there on mac an windows you will see PYTHON= the .emscripten config file even though nobody reads this or uses it. Its very confusing to have a setting like that that doesn't do anything.

We need the invent new functional FOO=bar settings for all tools, just to support making is_active() work as expected.

[sbc100]

I tired to mitigate the potential risk by building migration system from the old to new system. The idea is that maybe one day we can remove this.