Not really sure why this was done in the first place, but a kernel can be
forcibly killed and then asked to started up again, e.g. when restarting the
kernel.
Make less assumptions about where `point` is at when inserting results and
about the insertion context by relying only on
`org-babel-where-is-src-block-result` to move `point`. Centralize handling of
`point` to `jupyter-org--do-insert-result`.
* jupyter-org-client.el
(jupyter-org--first-result-context-p): Handle more cases.
(jupyter-org--append-stream-result-p): New function.
(jupyter-org--prepare-append-result): New function, abstracted from
`jupyter-org--insert-result`.
(jupyter-org--normalized-insertion-context): New function.
(jupyter-org--do-insert-result): New function.
(jupyter-org--insert-result): Move the bulk of the work to
`jupyter-org--do-insert-result`.
(jupyter-org--add-result): Use it.
Also consider indentation when inserting results. This was fairly easy to do
and didn't seem to warrant a separate commit.
* jupyter-org-client.el
(jupyter-org--insert-element): Remove. Update all callers.
(jupyter-org-delete-blank-line): New function.
(jupyter-org--delete-element): Use it.
(jupyter-org--stream-context-p): Consider indentation.
(jupyter-org--fixed-width-append): Remove. Update all callers.
(jupyter-org--append-to-fixed-width): Move example block promotion to
`jupyter-org--append-stream-result`.
(jupyter-org--append-to-example-block): New function.
(jupyter-org--append-stream-result): Use it.
(jupyter-org-indent-inserted-region): New macro.
(jupyter-org--insert-result): New method that splits the bulk of the old
function `jupyter-org--append-result` into a more manageable structure. Uses
new macro.
(jupyter-org--append-result): Remove it. Update all callers.
(jupyter-org--add-result): Refactor to consider above changes.
The completion at point functions are called several times in the course of a
single completion attempt so we get multiple chances to receive a completion
result. This means there is no need to wait for it to be received once and for
all on the first time completions are requested.
So that client subclasses can provide specialized behavior if needed.
Also adds the function `jupyter-eval-add-callbacks` so that client methods can
more easily obtain the default behavior of `jupyter-eval-string'.
