bug#66288: 29.1; Performance regression using pipe for subprocess

[Gregory Heytings]

> Therefore, one simple change is for xscheme.el to make the value of 
> read-process-output-max to 64KB.  Chris, can you test that, using the 
> unmodified 29.1 sources?

I'm not Chris, but that works indeed.

> A perhaps better change is the one below, which realizes that the fcntl 
> call was added to create_process for fixing bug#55737 so as to allow 
> _enlarging_ the pipe size when very large reads are required; it was 
> never meant to _reduce_ the default size of the pipe:

Indeed.