Move writing scheme details out of logic

[psvenk]

Currently, certain details that pertain to certain writing schemes are hard-coded into sanscript.js — for example, roman schemes are distinguished from Brahmic schemes by running addRomanScheme on schemes from a hard-coded list of roman schemes, and the Kolkata and itrans_dravidian romanizations are implemented in terms of IAST and ITRANS respectively which complicates the implementation of Sanscript and keeps scheme details in the otherwise scheme-agnostic code instead of in the scheme descriptions (code reference).

Moving this out of the logic eases maintenance of Sanscript implementations in multiple languages and more cleanly separates code from data, in the same vein as #13.