"uuid" is a primary key in "messages" but not in "messages_index",
the implication of that is very slow matching by UUID.

What can be done instead is matching messages_index.rowid to messages.rowid,
that is, an always-present clustered index.

This not only improves performance of full-text search but also of just
updating messages in any shape or form.

It now uses the data from the messages table instead of having a copy of each
message.

The message UUIDs are no longer part of the index.

If we don’t executePendingTransactions we might still access the overview fragment
while a replacement operation is in the works. This will lead to two
conversationfragments opening.

This reverts commit e528b9f5.

I was originally convinced by the argumentation (quote always in same place)
but testing this out for a while really seems to break 'last correct' for me.

I use that way more frequently that quote

- Add setting to prevent screenshots
- Enforce using FLAG_SECURE in onResume for each activity

Rewrite QuoteHelper to integrate French quotes logics. Also reallow QuoteChars not followed by whitespace as indicated in XEP-0393.

QUOTING_MAX_DEPTH=1 for transitory compatibility with older versions. QUOTE_MAX_DEPTH=7 for performance testing and hiding of a rerendering bug occuring when two adjacent messages are merged.