Relax cvxopt constraint on 1.2.1 to allow conda installation

Right now `requirements.txt` imposes an exact `cvxopt==1.2.1` version. This allows for a pip installation, but the `conda` package manager [only has version 1.2.0 available](https://anaconda.org/anaconda/cvxopt), which makes the installation conflict and potentially fail.

**Describe the solution you'd like**
Relax the installation requirement to `cvxopt>=1.2.0` or `cvxopt==1.2.*` to allow installation of DPPy in a conda environment without using `pip` for anything beside `DPPy` itself.

**Describe alternatives you've considered**
Other possibilities are:
- changing the constraint to `cvxopt==1.2.0`, and give up on the bugfixes of the point release
- open an issue asking continuum to update their package, can be done [here](https://github.com/ContinuumIO/anaconda-issues/issues)
- install cvxopt from [conda-forge](https://anaconda.org/search?q=cvxopt), the community-maintained conda repo. However as of now the package is at version 1.2.3
- install DPPy and dependencies only using pip in a ad-hoc environment to avoid conflicts

**Additional context**
This was the only requirement I could not meet using conda. Note that for generating the docs `sphinxcontrib-bibtex` is also not present in the default conda repo, but can be installed through conda-forge. Overall, it would be nice to have an easy installation path through conda, and eventually to be able to install DPPy directly using conda-forge.