Was writing `size` chars but the `hl` array was allocated as `rsize` chars.

With `'\` or `"...\` as the last characters on the line, `HL_STRING`
could be written beyond the allocated size of the `hl` array, causing
memory corruption and SEGV.

The test for a keyword would try to read the entire length of each keyword
from the line position even if the rest of the line was too short to
contain that keyword.
A PR was already submitted for this but the author closed it before
it could be merged.

The function `editorRowHasOpenComment()` used its own definition of a
closing multiline comment and duplicated the `hl_oc` member of `struct erow`.
Change to use the latter throughout and thus respect the selected syntax
definition.

del key used to behave just like backspace.
commit is based on d-e-h-i-o@799c743
but works on first char in file and doesnt crash program on last.

since python multiline string/comments start and end with the same
sequence, kilo is unable to figure out which is which once one removes
either one of the start or end markers of the section.
dealing with this would require a quite more elaborate engine than
currently implement, or doing a new syntax scan over the entire file.

for a cursor move that doesnt result in any scrolling, we now simply update
cursor position (and col/row in status bar) to remove the nasty flickering
caused by kilo on xterm when redrawing the entire screen everytime a cursor
is moved.
there's still a bug when moving cursor horizontally inside a long line, in
that it's not detected that the screen needs redraw.
i'll leave the fix to someone else as i'm done with this buggy editor.