Edit-distance weights need to be easily configurable

[aarppe]

As an improvement and generalization to #195, we need a non-computational way of manually specifying edit-distance weights by linguists etc..

In the first instance, this could be done context-independently, though on the longer term one would want to restrict certain weight-adjusted edits to certain contexts. The weights can be extracted from the regular expression that is specified for the weighted descriptive FST:

[
  ! Contraction of short /i/ marked with apostrophe 
  %' (->) i::0.0 ,,
  ! English interference
  d (->) t::0.0 ,, g (->) k::0.0 ,, j (->) c::0.0 ,,
  {iw} (->) {ow}::0.5 ,, {ow} (->) {iw}::0.5 ,,
  {tch} (->) c::0.0 ,, {ch} (->) c::0.0 ,, {ts} (->) c::0.5 ,,
  {ee} (->) [ i | î ]::0.0 ,, u (h) (->) a::0.0 ,,
  ! SRO-internal orthographical variation in hyphenation
  [..] (->) %-::0.5 ,,
  ! Cree-internal perception
  e (->) ê::0.0 ,,
  a (->) â::0.25,, i (->) î::0.25 ,, o (->) ô::0.25 ,,
  â (->) a::0.5,, î (->) i::0.5 ,, ô (->) o::0.5 ,,
  [..] (->) h::0.5 || [ a | â | ê | i | î | o | ô ] _ [ c | k | p | t ] ,,
  h (->) 0::0.75 || [ a | â | ê | i | î | o | ô ] _ [ c | k | p | t ]
] ;

Every other edit should have a standard weight of 1.0.

[aarppe]

The above code follows regular Xerox-style regular-expression syntax for rewrite rules (Karttunen & Beesley 2003).