This allows us to have newline not affect the wrap status of a line.
Now a lines wrapping status is changed only when the last cell
in the line is changed. This actually matches the behavior of many other
terminal emulators so is probably a good thing from a ecosystem
compatibility perspective.
The fish shell expects this weird behavior of newline not changing
wrapping status, for unknown reasons, which is the actual motivation for
doing all this work.
Fixes#5766
This ensures that there is no leftover line when the shell redraws the prompt,
by making the prompt a simple single character prompt before rewrapping.
Needed for output of hyperlinks, also more efficient, since avoids
malloc per line. Also fix pagerhist not having SGR reset at the start of
every line.
Split up the Cell structure into a CPUCell and a GPUCell. Only the
GPUCell part needs to be sent to the GPU. Should make kitty use even
less system resources and make a performance difference on systems where
the GPU bandwidth is constrained.
Also allows adding more CPU only data in the future without affecting
GPU bandwidth. For example, hyperlinks or more combining characters.
Instead of using heuristics to position the cursor after a resize, track
the position during re-wrapping and place the cursor at the re-wrapped
position. Fixes#242 (I hope)
This means that more text attributes can be added in the future (there
are 8 bits available) and there is no need to keep bit twiddling when
accessing the chars themselves. Also means less data needs to be sent to
the GPU. Cell size does not change since there were currently 2 unused
bytes because of alignment.
