Use --satisfied-skip-solve on conda install

[humitos]

Skip (do not upgrade) already satisfied dependencies that were
installed by the user when creating the environment, when installing
our own dependencies.

I'm building the images locally to do some QA before merging it.

Closes #3829
Related readthedocs/readthedocs-docker-images#102

[humitos]

This feature was introduced on conda 4.6.0 and we are running < 4.6 in production.

https://github.com/conda/conda/blob/master/CHANGELOG.md#improvements-6

So, we can't merge and deploy this PR until we upgrade all the images first. On the other hand, this can be added under a feature flag, but I think it's better to make a new deploy of the Docker images at this point.

[humitos]

OK. I found that this is still a problem for our use case :/

I did some tests with this repo (https://github.com/rtfd/test-builds/tree/conda-satisfied-skip-solve)

$ conda env create -f environment.yml
...

$ conda list | grep sphinx                                                                  
sphinx                    1.8.1                 py37_1000    conda-forge
sphinx-autorun            1.1.0                    pypi_0    pypi
sphinx_rtd_theme          0.4.3                      py_0    conda-forge
sphinxcontrib-websupport  1.1.0                      py_1    conda-forge
(condabuild) (miniconda3-latest) 

$ conda install --quiet --satisfied-skip-solve sphinx                                 

# All requested packages already installed.

Up to here, it works as I expected. Although, when there is at least one package that needs to be upgraded, it upgrades all the packages and we don't want that:

$ conda install --quiet --satisfied-skip-solve sphinx mock
Collecting package metadata: ...working... done
Solving environment: ...working... done

## Package Plan ##

  environment location: /home/humitos/.pyenv/versions/miniconda3-latest/envs/condabuild

  added / updated specs:
    - mock
    - sphinx


The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    certifi-2019.3.9           |           py37_0         155 KB
    mock-2.0.0                 |           py37_0         104 KB
    pbr-5.1.3                  |             py_0          74 KB


    sphinx-2.0.1               |             py_0         1.4 MB


    sphinxcontrib-applehelp-1.0.1|             py_0          29 KB
    sphinxcontrib-devhelp-1.0.1|             py_0          23 KB
    sphinxcontrib-htmlhelp-1.0.2|             py_0          28 KB
    sphinxcontrib-jsmath-1.0.1 |             py_0           8 KB
    sphinxcontrib-qthelp-1.0.2 |             py_0          26 KB
    sphinxcontrib-serializinghtml-1.1.3|             py_0          24 KB
    ------------------------------------------------------------
                                           Total:         1.9 MB

The following NEW packages will be INSTALLED:

  mock               pkgs/main/linux-64::mock-2.0.0-py37_0
  pbr                pkgs/main/noarch::pbr-5.1.3-py_0
  sphinxcontrib-app~ pkgs/main/noarch::sphinxcontrib-applehelp-1.0.1-py_0
  sphinxcontrib-dev~ pkgs/main/noarch::sphinxcontrib-devhelp-1.0.1-py_0
  sphinxcontrib-htm~ pkgs/main/noarch::sphinxcontrib-htmlhelp-1.0.2-py_0
  sphinxcontrib-jsm~ pkgs/main/noarch::sphinxcontrib-jsmath-1.0.1-py_0
  sphinxcontrib-qth~ pkgs/main/noarch::sphinxcontrib-qthelp-1.0.2-py_0
  sphinxcontrib-ser~ pkgs/main/noarch::sphinxcontrib-serializinghtml-1.1.3-py_0

The following packages will be UPDATED:

  sphinx             conda-forge/linux-64::sphinx-1.8.1-py~ --> pkgs/main/noarch::sphinx-2.0.1-py_0

The following packages will be SUPERSEDED by a higher-priority channel:

  ca-certificates    conda-forge::ca-certificates-2019.3.9~ --> pkgs/main::ca-certificates-2019.1.23-0
  certifi                                       conda-forge --> pkgs/main
  openssl            conda-forge::openssl-1.1.1b-h14c3975_1 --> pkgs/main::openssl-1.1.1b-h7b6447c_1


Proceed ([y]/n)? 

It will update sphinx to 2.0.1.

[humitos]

I just tried this with miniconda 4.6.14 and this still happens.

[humitos]

I'm closing this PR for now since we have a feature flag that helps us here. We can revisit in the future when conda behaves as we need.