improve FeatureRepository impl performance

[goekay]

the backing data structure for features was an ArrayList, which has O(n) worst-case
look-up performance. since the method findFeature(...) is frequently used, this has negative
implications on the performance of the library. A HashMap is better suited for feature look-up.

[coveralls]

Coverage increased (+0.3%) to 43.472% when pulling d1a4a83 on goekay:master into 2ee40bc on ChargeTimeEU:master.

[codecov-io]

Codecov Report

Merging #96 into master will increase coverage by 0.3%.
The diff coverage is 71.42%.

@@             Coverage Diff             @@
##             master      #96     +/-   ##
===========================================
+ Coverage     41.87%   42.17%   +0.3%     
- Complexity      876      882      +6     
===========================================
  Files           214      214             
  Lines          4227     4225      -2     
  Branches        438      435      -3     
===========================================
+ Hits           1770     1782     +12     
+ Misses         2351     2337     -14     
  Partials        106      106

[TVolden]

Thanks for your improvement, which makes good sense.

The code looks good. I only have one comment.

The tests could do with some better names to clearly state what they're testing and one assert per test. This will pay off when a failing test clearly states what it did and what it expected. I wrote this article about it https://github.com/ChargeTimeEU/Java-OCA-OCPP/wiki/Naming-unittest