Alignment: add TransformCelestialToTelescopeJD / TransformTelescopeToCelestialJD

[ckemper67]

Reviewer note

This PR changes a core alignment subsystem API (MathPlugin, MathPluginManagement)
that is used by multiple drivers inside this repo and by third-party drivers in
indi-3rdparty (notably indi-celestronaux and indi-eqmod). Would appreciate
architectural feedback before merge.

Summary

Refactors the INDI alignment subsystem so that the core transform methods no longer read
the system clock internally. Two new methods are added to MathPlugin that accept an
absolute Julian Date directly:

• TransformCelestialToTelescopeJD(RA, Dec, JulianDate, TDV)
• TransformTelescopeToCelestialJD(TDV, RA, Dec, JulianDate)

All built-in plugins (BasicMathPlugin / SVDMathPlugin, NearestMathPlugin,
SPKMathPlugin, DummyMathPlugin) implement the new methods as their primary
implementation. The old TransformCelestialToTelescope / TransformTelescopeToCelestial
methods become thin compat shims that resolve ln_get_julian_from_sys() + JulianOffset
and delegate to the new JD variants. MathPluginManagement gains matching *JD dispatch
methods; its old public methods are similarly shimmed.

Motivation

The old interface forced every plugin to call ln_get_julian_from_sys() internally.
This had two consequences:

1. Tests were non-deterministic. The test suite worked around this with a
   joff = fixedJD - ln_get_julian_from_sys() hack to re-anchor transforms to a fixed
   point in time. This was fragile and obscured intent.

2. Simulated time was impossible. Any future support for injected/simulated time
   (e.g. KStars time simulation) requires callers to supply the JD — the plugins cannot
   intercept ln_get_julian_from_sys() from the outside.

Backward compatibility

External plugins that have not yet adopted the new interface are unaffected at the
source and binary level. The new *JD methods are non-pure virtual. The base-class
default computes JulianOffset = JulianDate - ln_get_julian_from_sys() and calls the
old pure-virtual, so an unmodified external plugin recovers the correct absolute JD
when it adds ln_get_julian_from_sys() internally.

All existing drivers that call the old TransformCelestialToTelescope /
TransformTelescopeToCelestial interface — including skywatcherAPIMount,
indi-celestronaux, eqmod, telescope_simulator, astrotrac, temmadriver, dsc —
continue to work unchanged through the compat shims. Non-zero JulianOffset calls
(e.g. bracket tracking in skywatcherAPIMount and celestronaux) are correctly
handled: the shim resolves the absolute JD before dispatching.

Changes

libs/alignment/MathPlugin.h/.cpp

• Add TransformCelestialToTelescopeJD and TransformTelescopeToCelestialJD as
  non-pure virtual methods with default fallback implementations.

libs/alignment/BasicMathPlugin.h/.cpp, NearestMathPlugin, SPKMathPlugin, DummyMathPlugin

• Override the new *JD methods as the primary implementation (clock-free).
• Reduce the old Transform* overrides to one-liner shims.

libs/alignment/MathPluginManagement.h/.cpp

• Add public TransformCelestialToTelescopeJD / TransformTelescopeToCelestialJD
  dispatch methods (direct virtual call on pLoadedMathPlugin).
• Convert old public Transform* methods to compat shims.
• Remove now-dead pTransformCelestialToTelescope / pTransformTelescopeToCelestial
  function pointer members (dispatch goes through direct virtual calls).

test/alignment/test_alignment_plugins.cpp

• Remove the joff = fixedJD - ln_get_julian_from_sys() workaround from all test
  blocks; replace with direct *JD calls using static constexpr double fixedJD.
• Add two regression tests (LegacyPlugin_JD_Fallback_CelestialToTelescope and
  LegacyPlugin_JD_Fallback_TelescopeToCelestial) that verify the base-class fallback
  correctly converts an absolute JD to the offset expected by legacy external plugins.

Test results

[==========] 110 tests from 2 test suites ran.
[  PASSED  ] 110 tests.

Follow-up (not in this PR)

Phase 2 — Deprecate: Mark the old Transform* pure-virtual methods [[deprecated]]
to produce compiler warnings for external plugins that haven't migrated.

Phase 3 — Driver API: Thread a JD parameter (with default = ln_get_julian_from_sys())
through the six public methods of AlignmentSubsystemForDrivers
(SkyToTelescopeEquatorial, TelescopeEquatorialToSky, etc.) and update the internal
call sites to use *JD directly. Migrate direct MathPluginManagement callers
(skywatcherAPIMount, indi-celestronaux) to use *JD with explicit JD values.

Phase 4 — Remove: Once no callers use the old interface, make *JD pure virtual and
remove the compat layer entirely.

[knro]

Great work! I agree this would make all the testing and simulation deterministic now. You can mark the shim functions as deprecated. At the same time, another PR for all affected drivers should be prepared to migrate them to the new functions.

[ckemper67]

Yes I will send the follow-up PRs once we have this submitted. Do you want the depreciation as part of this pull request or in a follow up as well?

[knro]

Sure you can add it to this PR as well.

[ckemper67]

On second thought, I would prefer to combine the deprecation declaration with the PR that fixes the drivers. That way we don't introduce deprecation warnings/errors into the build. What do you think?

#2388 is the followup. I based it on master, so it contains this PR as well. The last commit are the deprecation changes and the cleanups.