Use tagged versions of tomopy and dxchange for the tomopy reconstruction example

[mrakitin]

A new version of tomopy can be installed with pip install .. See #97 which uses this new approach and installs tomopy from the master branch. Once the new version of tomopy is tagged/released (most likely https://github.com/tomopy/tomopy/releases/tag/1.2.2), we should use the tagged version and install it with:

pip install git+https://github.com/tomopy/tomopy@v1.2.2

Currently, we install the dxchange package from the master branch of their repo: https://github.com/NSLS-II/docs/blob/c1bd5347d63ea6133feba3545b9ce46536ef98ce/requirements.txt#L25 Once the installer of dxchange is updated (see my issue https://github.com/data-exchange/dxchange/issues/81), we will be able to install it with:

pip install git+https://github.com/data-exchange/dxchange@v0.1.6

[mrakitin]

There are two aspect of the issue:

• The dxchange package needs some improvements to rely on it to automatically install the dependencies including pandas. See https://github.com/data-exchange/dxchange/issues/81 for more details. So, at the next release of dxchange the tagged versions won't help much to avoid missing packages. A proper pip-installable package will solve the issue. Meanwhile we can just keep the dependency in our requirements.txt.
• Using the tagged version to have a frozen set of files, so we have less failures of building the docs.

[mrakitin]

I updated the issue description and the title to have this change implemented for both tomopy and dxchange.