Docs: Added a step-by-step documentation for installing and building LPython

[kmr-srbh]

This PR adds the step-by-step installation instructions for setting up and building LPython on Linux, Windows, macOS and Windows with WSL. It overrides all of the dummy LFortran data previously present.

The steps for setting up the project on Windows with WSL are taken from #1501, courtesy of @faze-geek.

[certik]

I think this looks good. Thank you for improving the installation instructions!

[Shaikh-Ubaid]

Last time when I was setting up LFortran on Linux, I noted the extra-dependencies that were needed to be installed after setting up the conda environment:

1. build-essential (for gcc, g++ and make)
2. binutils-dev (for error: Could not find BFD and also for stacktraces support)
3. zstd-static=1.5.5 (for zstd-static not found during link)
4. zlib1g-dev (to fix ld error -lz not found)
5. clang (as it is used by lfortran for linking when compiling user source code)

I installed 1, 2, 4, and 5 using sudo apt-get install.
3 should be installed via conda.

Please note the above are for LFortran while setting up in Ubuntu/Linux operation system.

[certik]

Yes. For everything else we should use the conda environment. Especially important for Bison, which only works with version 3.4 and often fails with newer versions.

[Shaikh-Ubaid]

Also from https://lfortran.zulipchat.com/#narrow/stream/197339-general/topic/Building.20Lfortran/near/420682935, it seems zlib1g-dev should be installed via conda. So, summarizing the above, I think the steps to setup an LCompiler on Linux can be:

• Step 0 (Prerequisite): Download and install conda if not already installed.
• Step 1: Create the conda environment as per the appropriate environment file in the project
• Step2: sudo apt-get install build-essentials binutils-dev clang

So, seems like just two steps.

Notes:

• we should update this environment files to also include zstd-static and zlib1g-dev
• we might be able to eliminate installing clang, but I think for now clang has been hard coded to be used for linking in LFortran. I think LFortran should ideally use the compiler specified by LFortran_CC for linking when compiling user source code.

[kmr-srbh]

Also from https://lfortran.zulipchat.com/#narrow/stream/197339-general/topic/Building.20Lfortran/near/420682935, it seems zlib1g-dev should be installed via conda. So, summarizing the above, I think the steps to setup an LCompiler on Linux can be:

* Step 0 (Prerequisite): Download and install conda if not already installed.

* Step 1: Create the conda environment as per the appropriate environment file in the project

* Step2: `sudo apt-get install build-essentials binutils-dev clang`

So, seems like just two steps.

Notes:

* we should update this environment files to also include `zstd-static` and `zlib1g-dev`

* we might be able to eliminate installing clang, but I think for now clang has been hard coded to be used for linking in LFortran. I think LFortran should ideally use the compiler specified by `LFortran_CC` for linking when compiling user source code.

I am making the changes. 👍 Thank you very much for the help!🙂

[Shaikh-Ubaid]

we should update this environment files to also include zstd-static and zlib1g-dev

It seems like zlib1g-dev is not present on conda-forge. So we need to be careful here.

@certik do you have any suggestions for installing zlib1g-dev. It seems it is used only for c/c++ zlib header files.

[kmr-srbh]

* we should update this environment files to also include `zstd-static` and `zlib1g-dev`

I am adding the packages. As @Shaikh-Ubaid pointed out, zlib1g-dev is not present on conda-forge. A global installation works. Is it fine?

[Shaikh-Ubaid]

Is it fine?

Seem fine to me for now. We can update later as we figure out better approaches.

[Shaikh-Ubaid]

Also, I noticed that you are working in main branch directly (for example this PR or the commit above). It is usually suggested to use separate branches to submit developmental code. You use main branch to pull in latest changes from the remote repository. You use other branches to work on developmental/experimental code and later submit as a PR when ready.

[kmr-srbh]

You are right @Shaikh-Ubaid. To get started quickly, I had started work on the default branch. I have created new work-specific branches and am working on them now. 🙂