Reset interpreter session and reuse the interpreter (isolation level)

[eldipa]

Describe the feature you'd like
For each file, byexample spawns an interpreter to run the examples and it will shutdown after processing the file.

If multiple files have examples of the same language, byexample will still spawn and shutdown the interpreter for each file.

This is quite slow.

Some interpreters have the concept of sessions: independent execution sessions with a clean environment.

We could use it to reset the session instead of spawning another interpreter which it should be much faster.

This opens several isolation levels:

• process: restart each interpreter, always
• session: restart the session if supported, restart the interpreter otherwise
• cleanup: within the same session, perform some actions to clean up it.
• none: do not do anything, it is the user that must code her examples correctly. byexample must still do some clean up like resetting the clipboard (isolation == none is not a way to concatenate files)

[eldipa]

Spawning and stopping an interpreter per file ensures that the interpreter is closed. If the next file to exec does not use that interpreter it is okay.
But if we impl this "reset" feature, an interpreter will survive the exec of one file even if the next file does not require it.

We need to ensure that the interpreter is shutdown if no other file will require it.

Offtopic: it would be interesting to make byexample to schedule the files to exec based on the interpreters that requires. When we run them in parallel it would be cool to balance the load: mix files that require less or more interpreters so we don't spawn a lot of interpreters for some moments and then run very few in others.