mapcairo.c: avoid double scaling of SVG symbols (fixes: #7430) - #7583

This fails two CI checks, because the use of the deprecated function svg_handle_render_cairo() (which is the base of this workaround).

• maybe I can add exemption from throwing a build error?
• else proper a solution with rsvg_handle_render_document() might be needed

@johanez there are actually 2 different failing errors in your tests (click on each of the 2 runs, and you will see what I mean). For example: https://github.com/MapServer/MapServer/actions/runs/29905134730/job/88913132480?pr=7583

@jmckenna I found another possible solution, using the non-deprecated rsvg_handle_render_document() and not using cairo_scale() for rsvg rasterization, as the scaling is already done by setting the viewport.

@johanez can you join the discussion in #7430 to propose your change / give feedback for the other proposed change?

@johanez please also update the title of this pull request, thanks.

I updated the test files. As svg rendering was not working correctly, the coordinates used in the test point-identify with anchor test were not correct.

Update from my side: I used the solution suggested by @bhark, as it more easily allows to be compatibel with older librsvg versions (<2.43) as well.

Some test still fail, I think mostly because the expected results are made with broken svg

However, it's quite a task for me to go through this (skill and capacity wise, I have a very un-efficient test setup now), so if anyone want to have a look, more than welcome!

@johanez thanks. At this point, the issue isn't fixing the tests in this pull request, the issue is that people in #7430 have confirmed a different fix, than this one. I really want everyone to agree on one fix...I'll give this a few days more to await everyone's feedback here. Thanks for pushing this forward @johanez very appreciated!

@jmckenna The fix implemented here is the same as i suggested in #7430, and seems functional to me. #7462 just adds the guard, which is already in this PR.

With regards to the failing tests @johanez the original commit that introduced this issue also regenerated the reference pictures, so they're currently checking the correct rendering against a faulty baseline. I think we just need to regenerate the references (all 6) and call it a day. I'll open a PR on @johanez fork if i get the time later today.

@johanez @bhark I've updated the expected test results here, and the rendering looks a lot better (see 87144bb )

Thanks @jmckenna and @bhark . I indeed changed it into using the same fix as @bhark suggested, as it's simpler to keep the older librsvg version guard.

Looks like this is ready to merge and great if you backport it to 8.6.6!

thanks again @johanez and @bhark !