This PR adds a mode for automatic opening of brackets, braces and braces.

This follows a strange idea that appeared on Saturday the 13th, following the other strange idea in #14077.

The logic used is as follows:

• if the mode is enabled
• and if the cell is only one line long
• and if there are unmatched [{()}]
• and if it is enough to add opening symbols at the beginning and/or closing symbols at the end

then the line is modified.

Technical points:

• I chose not to use prefilters, but input transformers for this, because prefilters work with the check/handle logic, which does not allow to have at the same time the addition of symbols and the use of a magic (like %time for example).
• the TransformerManager didn't have a member allowing to come back to the shell, and it's essential for me to access the shell config and to be able to display the modified line (when there is a modification)
• the transformations had no way to have additional arguments and only received the lines. I had to add kwargs to be able to add arguments that transformations are free to use or not.

Demo:

Python 3.11.2 (main, Mar 13 2023, 12:18:29) [GCC 12.2.0]
Type 'copyright', 'credits' or 'license' for more information
IPython 8.14.0.dev -- An enhanced Interactive Python. Type '?' for help.

In [1]: 1+1)/2
  Cell In[1], line 1
    1+1)/2
       ^
SyntaxError: unmatched ')'


In [2]: %autobalance
Autobalance enabled

In [3]: 1+1)/2
------> (1+1)/2
Out[3]: 1.0

In [4]: 1+1)/(3+4
------> (1+1)/(3+4)
Out[4]: 0.2857142857142857

In [5]: i+1)/2 for i in range(5)]+[1,2
------> [(i+1)/2 for i in range(5)]+[1,2]
Out[5]: [0.5, 1.0, 1.5, 2.0, 2.5, 1, 2]

In [6]: my_values = i+1)/2 for i in range(5)]+[1,2
------> my_values = [(i+1)/2 for i in range(5)]+[1,2]

In [7]: %config TerminalInteractiveShell.show_rewritten_input = False

In [8]: 1+1)/(3+4
Out[8]: 0.2857142857142857

This is my first time diving into IPython code (and contributing to it), I may have missed a lot things. I place this PR in draft for two reasons:

• the changes made to the TransformerManager logic (with shell passing and kwargs on transformations) seem a bit intrusive to me, and discussing them seems like the best option.
• I have not yet written the specific tests for this new behavior.

Finally, I think it is important to mention the adapted xkcd.