switch base environment to requirements file

[minrk]

requirements file is easier to work with and update than inlined strings

In the future, we should be doing pip freeze and using the frozen result, like we do (with conda) in repo2docker because pinning only direct dependencies is a way to guarantee broken environments in the future (see the pinning-down of tornado as one example).

Several dependencies are updated in the process (jupyterlab 1.1, notebook 6.0, etc.)

[yuvipanda]

@minrk totally agree re: freeze.

Failing with:

Ran /opt/tljh/user/bin/jupyter labextension install @jupyterlab/hub-extension @jupyter-widgets/jupyterlab-manager with exit code 1
-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�Building jupyterlab assets (build:prod:minimize)
-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�|�\�-�/�An error occured.
RuntimeError: JupyterLab failed to build
See the log file for details:  /tmp/jupyterlab-debug-vdawix52.log

With 1.1 we don't need hub-extension either.

[GeorgianaElena]

I ran into this problem (jupyterlab failing to build) on my TLJH deployment on DO when I upgraded jupyterlab manually from 0.35.4 to 1.2.5.
From the logs, it looked like the failure was related to jupyterlab/jupyterlab#7175. After increasing the memory through the env var NODE_OPTIONS, the build went through.

I tried to solve the failing tests in this PR it in the same manner but with no luck :( Either there's something missing or the tests are failing from a whole different reason.

[yuvipanda]

@GeorgianaElena I wonder if this needs a rebase - we fixed some other errors in master after this PR I think.

[lwbayes]

Another reason for flaky build is that Webpack minify process is resource intensive, which might cause flaky build on pipelines with smaller machines. add --minimize=False to the labextension install seems to solve the issue for me. Will raise a PR later

[GeorgianaElena]

I've rebased the PR and now it contains @letianw91 's fix 🚀

[yuvipanda]

Awesome, ty @GeorgianaElena!