v0.5: Scoring and evaluation in one script

Added candidate evaluation to "gismu_score.py".

Additional rules to exclude forbidden or ambiguous letter sequences.

Deduplication and serialization of candidates are now options which are
disabled by default.

v0.4: Pure-python LCS, custom weights and shapes

Replaced mlpy dependency with pure-python LCS implementation
for easier installation and compatibility with additional
python implementations.

Candidate generation now defaults to only use letters from
input words.

Added options to customize language weights, gismu shapes
and alphabet, and number of scoring threads.

v0.3: Correct scores per source-word length

Per CLL 4.14.4: "The scores were divided by the length of the source-language
word in its Lojbanized form"

v0.2: Uses mlpy LCS (Longest Common Subsequence)

Initial commit as downloaded from lojban.org

Python implementation of gismu scoring algorithm by Arnt Richard Johansen:

  http://www.lojban.org/tiki/gismu+scoring+algorithm