Build tool update

[jasonb5]

Adds some new cli arguments:

• conda_env: Specifies the conda environment to use.
• extra_channels: List of channels to set in the condarc.
• ignore_conda_missmatch: If true --no-check-uptodate is set when calling conda smithy rerender, this handles a situation when conda-smithy or conda-forge-pinning are not sync in the environment.
• conda_rc: Path to a file that will be used to configure conda.

[jasonb5]

Added a few more things.

• If a repo does not have a feedstock then we create a fake one in workdir and run conda smithy there. This prevents creating a dirty repo with conda smithys auto commits.
• Adds a flag to use a local repo, this will replace the source section in the recipe to use the local directory rather than a git commit.

[muryanto1]

testing it with cdat_info (which is a noarch, and conda-forge pkg), conda-rerender and conda-build passed, and the build creates
/home/circleci/project/linux_build/miniconda/conda-bld/noarch/cdat_info-8.2.2020.06.05.23.39.gcac191d-pyh9f0ad1d_0.tar.bz2

but when we try to do upload, we did: conda build --output to get the package name, and it is returning: cdat_info-8.2.2020.06.05.23.39.gcac191d-py_0.tar.bz2

[muryanto1]

The reason I had clone_repo() here is because doing rerendering will overwrite .circleci/config.yml.
The repo will be cloned to /<repo_name>.
If I accidentally set to the parent directory of my repo where I ran Makefile from, clone_repo() will overwrite my repo.
so it is better to put a check somewhere that should not be the same as the parent directory of the repo where I ran 'make' from.

[muryanto1]

This comments apply to both build_in_local_repo() and build_in_local_feedstock().
Currently when we do conda build, it gets the branch of the repo, which means if I do local changes, I have to commit to the branch and push it to github. Is there any way we can make 'conda build ..' to just use whatever we have in the work repo?