split cursor bugs
There are cases where pango computes the wrong positions for strong and weak cursors.
One example is:
ab<aleph><beth>|
Visually, this appears as
abאב|
Both cursor positions are at the end, but typing a gimel or c will lead to:
abאבג
abאבc
So arguably, the weak cursor should have been at visual position 2 and the strong cursor at visual position 4.
A more confusing case arises with deeper embedding:
ab<aleph><beth>12|c
Visually, this appears as
ab12|אב|c
with the strong cursor at visual position 4 and the weak cursor at position 6.
Yet, typing an gimel or a leads to:
abאב12גc
abאב12ac
So, the strong insertion happens where the weak cursor was displayed, while the weak insertion happens somewhere else entirely.
Just for fun value, typing a 3 inserts at the place where the strong cursor was displayed, so there's actually 3 different insertion positions here.