Jupyter-conda

Changes:
- Available package cache file is now writable for everybody to avoid trouble in multi-user context. 25
- Add update environment from file through REST endpoint [PATCH /environments/<env>](https://github.com/fcollonval/jupyter_conda/blob/8f6c17da68a50e19971a0a7974c1d0e8420c62ed/jupyter_conda/handlers.pyL265) 26
- Switch to newer Python syntax `async`-`await`
- To improve UI reactivity in Jupyterlab:
- Long running task can now be cancelled 32
- The available packages list is used to find updatable package. `conda update --dry-run --all` is not used any longer. But it is still used if the user request updating all possible packages.

- Add ability to specify kernel companions; i.e. check that if some packages are installed in a
kernel, they must respect a certain version range. Companions can be specified through user
settings.
- `IEnvironmentManager.getPackageManager()` returns always the same `Conda.IPackageManager`
otherwise signaling package operations would have been meaningless.
- Request environment list access now `whitelist`=0 or 1 query arguments. If 1, the environment
list is filtered to respect `KernelSpecManager.whitelist`. Default is 0, but it could be modified
in user settings.
- Small UI tweaks

* Rework the server/client API to be more REST
* Cache available packages list in temp directory
* Improve greatly the coverage for the server extension
* JupyterLab extension only:
* Allow the user to change the proposed environment when creating one from scratch
* Add signals to handle environnements and packages changes (see labextension\src\__tests__\services.spec.ts)
* Improve the UI reactivity by using react-virtualized for the packages list
* Improve the look and feel
* Allow to choose the version to install
* Available packages truncation has been removed.