* Fix compilation with non glibc.

_SC_NPROCESSORS_ONLN is non standard as defined on https://man7.org/linux/man-pages/man3/sysconf.3.html
It seems to only be on glibc, cf https://www.gnu.org/software/libc/manual/html_node/Processor-Resources.html

* Fix to defined(_SC_NPROCESSORS_ONLN)

3rdparty/libjpeg-turbo/src/jpeglib.h:
`* NB: you must set up the error-manager BEFORE calling jpeg_create_xxx.`

It was previously not clear that the extension should include the period, [and I am not the only person that encountered this problem](https://stackoverflow.com/questions/4254460/opencv-could-not-find-encoder-for-the-specified-extension).

I assume that this is true for all supported extensions, but I haven't checked the source.

In test_imgproc.js, the test_filter suite's last test assigns a variable
to `size` without declaring it with `let`, polluting the global scope.
This commit adds `let` to the statement, so that the variable is scoped
to the test block.

GTest filtered list for 3.4 branch

If there will be measurement before the next predict, `statePost` would be assigned to updated value. So I guess these steps are meant to handle when no measurement and KF only do the predict step.
```cpp
statePre.copyTo(statePost);
errorCovPre.copyTo(errorCovPost);
```

…_linux_3.4

Workflow to use a self-hosted Github runner on Linux x86_64 for 3.4 branch

* Changed a workflow to use a self-hosted Github runner for 3.4 branch

…pression cases for smaller files.

Minor documentation fixes for geometric image transformations

…-only

TIFFTAG_PREDICTOR should not be applied unless LZW, DEFLATE or ADOBE_DEFLATE

Workflow to build and test on ARM64 for 3.4 branch

* add ARM64 workflow for 3.4

* add opencv_contrib fork check

* rename RET to OPENCV_EXTRA_FORK

…ndows_actions

Documentation Fix: Python and Java switched in Explanation