Sanskrit-parser

Fixes Issue 184 - nissAraH is now split correctly.

Enable the use of `sqlalchemy > 1.4` (Issue 177 ) through an edit of `sanskrit_data.db`

**Full Changelog**: https://github.com/kmadathil/sanskrit_parser/compare/v0.2.4...v0.2.5

Cleaner usage of indic_transliteration.sanscript

UI change - front and back ends use strict_io=True for most cases. Javascript handles visarga display.

Cleaner usage of `indic_transliteration.sanscript`

UI change - front and back ends use `strict_io=True` for most cases. Javascript handles visarga display.

Adding a "Strict Input Mode" switch in the UI.
By default, strict mode is on, input is interpreted as is.
If this is unchecked
1. Lazy anusvAras are corrected (शंकरः ->‌शङ्करः)
2. End-of-string anusVaras are corrected to म्
3. ISO15919 is forced to IAST with a few translations
4. Some common encoding errors are fixed (: for H or visarga, for example).

Minor fixes