Validate CodeMirror input as valid SGML/NLP compatible

[amir-zeldes]

Invalid SGML can break the spreadsheet conversion, for example:

<lb>ⲧⲱⲛ_·>_ϫⲉ|ⲟⲩ|ⲣⲱ</lb>

Additionally, some NLP components expect specific constraints on data. For Coptic Scriptorium, element content (not inside attributes) may not contain - or |, since the NLP chain is expecting no pre-tokenization. The latter problem could potentially be solved in the NLP tools, but users would be unaware of what is probably an error.

Possible solutions:

1. Validate SGML to rule out unescaped >, < in element content (probably in JS in client, maybe highlighting in CodeMirror?)
2. Make an easily configurable script that also checks for specific characters in element content, configurable per instance, or via the validations table