Make idris-compiler-notes-mode derived from special-mode and

[keram]

align code structure of idris-list-compiler-notes with idris-hole-list-show.

Also includes few small improvements (see commits for details)

Why:
To improve maintainability.
This PR may also fix issue with company-mode idris-community/idris2-mode#36 or atleast get us closer as it removes unnecessary calls to pop-to-buffer/display-buffer .

Next steps:

• remove now unused variable idris-stay-in-current-window-on-compiler-error
• ensure that the notes buffer is not displayed in the same window as the loaded file when first connection to idris is created

[jfdm]

Several jobs failed. I am rerunning them to see if the issue is caching or something else entirely.

[keram]

Several jobs failed. I am rerunning them to see if the issue is caching or something else entirely.

I may try rebuild my version of Idris2 to check if something changed there by any chance.
So far haven't been able to reproduce any of these CI failures with Idris 2, version 0.6.0-f443723f4.
One interesting discovery I made this morning on M1 Mac is that the tests that get "randomly" stuck
at "Preparing compiler note tree..." from:

(defun idris-list-compiler-notes ()
  "Show the compiler notes in tree view."
  (interactive)
  (with-temp-message "Preparing compiler note tree..."
    (idris-compiler-notes-list-show (reverse idris-raw-warnings))))

start passing consistently after slight change (printing the idris-raw-warnings in debug message) 🤯

(defun idris-list-compiler-notes ()
  "Show the compiler notes in tree view."
  (interactive)
  (with-temp-message "Preparing compiler note tree..."
    (message "-t- idris-list-compiler-notes 0 %s" idris-raw-warnings)
    (idris-compiler-notes-list-show (reverse idris-raw-warnings))))