Use marisa_trie for storing root word corpus

Storing the root word corpus on a linear list is the least efficient way
to store it. Here, I switched it to use marisa_trie - a memory efficient
Trie datastructure based on marisa-trie C++ library. It is expected to
provide lesser space and time complexity.

.travis.yml

@@ -4,6 +4,7 @@ python:
   - "3.4"
   - "3.5"
 install:
+  - pip install -r requirements.txt
   - pip install -r test-requirements.txt
 script: make travis
 after_success: coveralls

libindic/stemmer/__init__.py

@@ -23,6 +23,8 @@
 
 import os
 
+import marisa_trie
+
 
 class Malayalam:
     """
@@ -38,10 +40,11 @@ def __init__(self):
         self.dictionary = self.dictionary_file.readlines()
         self.dictionary_file.close()
         try:
-            self.dictionary = [x.strip().decode('utf-8')
-                               for x in self.dictionary]
+            self.dictionary = marisa_trie.Trie([x.strip().decode('utf-8')
+                                                for x in self.dictionary])
         except:
-            self.dictionary = [x.strip() for x in self.dictionary]
+            self.dictionary = marisa_trie.Trie(
+                [x.strip() for x in self.dictionary])
 
     def singleencode(self, word):
         '''

requirements.txt

@@ -0,0 +1 @@
+marisa_trie

tox.ini

@@ -4,15 +4,17 @@
 # and then run "tox" from this directory.
 
 [tox]
-envlist = py27, py35, pypy, pep8
+envlist = py27, py35, pep8
 
 [testenv]
 commands = {envpython} setup.py test
 deps =
+    -rrequirements.txt
     -rtest-requirements.txt
 
 [testenv:pep8]
 deps=
+  -rrequirements.txt
   -rtest-requirements.txt
 commands=
   flake8 libindic