RISC-V: modified PR scripts

[mshabunin]

• Added single workflow for precommit RISC-V builds for 4.x and 5.x branches (newer approach with less duplicated code)
• Added build variant with HAL=OFF
• Fixed issue with anonymous git clone via https

TODO:

• disabled build with contrib, should we enable it?
• temporarily disabled linux and windows workflow checks - re-enable before merge

NOTE: github token propagation is needed to avoid failures of anonymous git clone via https, permissions set to minimum

@asmorkalov , some tests are failing, should we disable them for now or can someone take a look?

4.x With HAL=OFF

    [  FAILED  ] Core_Array.expressions
    [  FAILED  ] Core_SVD.accuracy
    [  FAILED  ] Core_SVD.flt
    [  FAILED  ] Core_Invert.small
    [  FAILED  ] Core_SolveLinearSystem.accuracy
    [  FAILED  ] Core_QR_Solver.accuracy64f
    [  FAILED  ] Core_Solve.regression_11888
    [  FAILED  ] Core_Solve.Matx_4_4

    [  FAILED  ] fitEllipse.small
    [  FAILED  ] Imgproc_FitEllipse_HorizontalLine.accuracy
    [  FAILED  ] Imgproc_getPerspectiveTransform.issue_26916
    [  FAILED  ] fitEllipse_Modes.accuracy/0, where GetParam() = 13
    [  FAILED  ] fitEllipse_Modes.accuracy/1, where GetParam() = 12

5.x With HAL=OFF

    [  FAILED  ] Core_DotProduct.accuracy (864 ms)
    maybe more...

[mshabunin]

@asmorkalov , for some reason Win photo test on 4.x branch consistently fails for me, even though I did not change anything in that job

    [ RUN      ] Photo_CalibrateDebevec.regression
    C:\GHA-OCV-6\_work\ci-gha-workflow\ci-gha-workflow\opencv\modules\photo\test\test_hdr.cpp(292): error: Expected: (max) < (0.15), actual: 0.214742 vs 0.15
    [  FAILED  ] Photo_CalibrateDebevec.regression (1401 ms)

Should we fix or disable this test?