fix dependencies doc and add pointer to v0.5.0 for earlier Python versions (#298)

Author: bellet

README.rst

@@ -20,7 +20,8 @@ metric-learn contains efficient Python implementations of several popular superv
 
 **Dependencies**
 
--  Python 3.6+
+-  Python 3.6+ (the last version supporting Python 2 and Python 3.5 was
+   `v0.5.0 <https://pypi.org/project/metric-learn/0.5.0/>`_)
 -  numpy, scipy, scikit-learn>=0.20.3
 
 **Optional dependencies**

doc/getting_started.rst

@@ -17,15 +17,16 @@ metric-learn can be installed in either of the following ways:
 
 **Dependencies**
 
--  Python 2.7+, 3.4+
--  numpy, scipy, scikit-learn>=0.20.3
+- Python 3.6+ (the last version supporting Python 2 and Python 3.5 was
+  `v0.5.0 <https://pypi.org/project/metric-learn/0.5.0/>`_)
+- numpy, scipy, scikit-learn>=0.20.3
 
 **Optional dependencies**
 
 - For SDML, using skggm will allow the algorithm to solve problematic cases
   (install from commit `a0ed406 <https://github.com/skggm/skggm/commit/a0ed406586c4364ea3297a658f415e13b5cbdaf8>`_).
   ``pip install 'git+https://github.com/skggm/skggm.git@a0ed406586c4364ea3297a658f415e13b5cbdaf8'`` to install the required version of skggm from GitHub.
--  For running the examples only: matplotlib
+- For running the examples only: matplotlib
 
 Quick start
 ===========

doc/introduction.rst

@@ -96,7 +96,7 @@ examples (for code illustrating some of these use-cases, see the
   metric learning provides a way to bias the clusters found by algorithms like
   K-Means towards the intended semantics.
 - Information retrieval: the learned metric can be used to retrieve the
-  elements of a database that are semantically closer to a query element.
+  elements of a database that are semantically closest to a query element.
 - Dimensionality reduction: metric learning may be seen as a way to reduce the
   data dimension in a (weakly) supervised setting.
 - More generally, the learned transformation :math:`L` can be used to project