Consider GitHub workflows to monitor failing installation processes

[ras44]

Describe the bug
While troubleshooting installation issues, I ended up adding some GitHub workflows to make each of the installation options in the README a reproducible process.

These include:

• Install from conda-forge action
• Install with the conda virtual environment action
• Install from PyPI action
• Install from source action

Some of the installation processes described in the README at: https://github.com/ras44/causalml#installation
are failing.

Build/install issues are mentioned in #624, #609, #593, #533, #534, and it may help user experience to have a reproducible workflow that illustrates they are building correctly (and how to build them).

To Reproduce
See the github workflows linked to above to reproduce.

Expected behavior
Following install instructions should create prod/dev environments without issues.

Have the README instructions for:

• Install from conda-forge
• Install with the conda virtual environment
• Install from PyPI
• Install from source

match a reproducible GitHub workflows that are passing, to build user confidence and avoid abandonment.

Environment (please complete the following information):

• OS: [Windows, Ubuntu]
• Python Version: [3.7, 3.8, 3.9]

Additional context

Does it sound like a reasonable plan to build GitHub actions that reproduce these install scenarios?

The idea is that they should all pass, and that the README instructions can then either reference them or copy them for the end user to use in a prod or dev setting.

[ras44]

Currently, all 6 py x tf conda envs build successfully, but tests fail:

https://github.com/ras44/causalml/actions/runs/5546320526/jobs/10126438492

Building from source for py [3.6, 3.7, 3.8] using the dependencies defined in pyproject.toml succeeds and tests succeed:

https://github.com/ras44/causalml/actions/runs/5546201960