[Oracle] Fix MultiSurface with straight polygon and take into account orientation when writing polygon rings

[rouault]

When fixing warnings of #33930, I
looked at the code that transforms a QGIS geometry of type MultiSurface
to a Oracle geometry, and it appeared quite convoluted/risky (variables
being reaffected with values of other variables), and not
being able to deal with straight Polygon in MultiSurface.
The reverse situation (Oracle MultiSurface to QGIS MultiSurface) had the
same issue as well.

The MultiCurve/CompoundCurve code has been modified similarly. There was
no real bug. Just a sub-optimal behaviour on reading of MultiCurve
from Oracle, where all parts where promoted to CompoundCurve, even when
not strictly needed.

Also fixes polygon writing by reverse rings when needed to honour the winding order of Oracle polygons (#29085)

Note: If backporting is wished, #33930 will have to be backported first

[nyalldawson]

@troopa81 can you review please?

[troopa81]

great work! looks ok to me, I just have one question for rectangle 3D

[troopa81]

It's all good for me, LGTM

[rouault]

@troopa81
OK, I see that in #33976 you propose to backport some functionality touching code related to the one of this PR. I'm not completely clear if all functionality that is in master related to Oracle geometry handling is backported per #33976 and thus if this PR be able to be backported ? (not critical to me for it to be backported. just assessing if it is doable)

[troopa81]

@rouault You won't be able to backport everything indeed because Curve/MultiCurve/MultiSurface exists only in master. All the code to reverse rings to force them to be in the Oracle appropriate order could be backported but I'm not sure it should be.

It was a limitation and you remove it, it's kind of a feature to me.

[rouault]

It was a limitation and you remove it, it's kind of a feature to me.

Fair enough. Keeping this as master-only material.