[Documentation] Include all assumptions made for index definition

[dariok]

/db/config/db/apps/edoc/data/collection.xconf (from config/data/collection.xconf) has to make a few assumptions about how the text is organised, e.g.

• that ab is not indexed when a descendant of text
• that del is not important (i.e., deletions are not indexed, be they standalone or as part of a subst – as subst may be inline, an inline subst would thus always resolve to the “add” version)
• or that choice (with sic and corr, or orig and reg) always works at word level and not on a character level (as <choice><sic>c</sic><corr>k</corr></choice>alender would be indexed as ckalender – the alternative would be to only use corr and reg, as in subst above)
• what an “information bearing element” is (text in general, and e.g. p within the text)
• which elements are indexed separately from the “information bearing” elements (e.g., head)