Allow micro simulations to pass local data for adaptivity calculation only

[AdityaGupta716]

Summary

Closes #25

Micro simulations can now return data from solve() (and optionally initialize()) that is used only for the adaptivity similarity distance calculation and is never sent to the macro simulation via preCICE.

This is useful when a micro simulation has internal state data (e.g. a phase indicator, convergence metric, or any other state variable) that characterizes similarity between simulations but should not be communicated to the macro side.

Usage

Add a local_data key to adaptivity_settings in the Micro Manager JSON config:

"adaptivity_settings": {
    "type": "local",
    "data": ["Macro-Scalar-Data"],
    "local_data": {"internal-state": "scalar"},
    ...
}

The micro simulation's solve() method includes the local data key in its returned dict:

def solve(self, macro_data, dt):
    return {
        "Micro-Scalar-Data": ...,   # sent to macro simulation as usual
        "internal-state": 42.0,     # used only for adaptivity, never sent to macro
    }

Changes

• micro_manager/config.py: reads optional local_data from adaptivity_settings, exposes it via get_local_data_for_adaptivity()
• micro_manager/micro_manager.py: initializes local data arrays, collects local data from solve(), initialize(), and lazy initialization paths, strips local data keys from output before writing to preCICE
• tests/unit/test_micro_manager.py: new test verifying local data is collected for adaptivity but absent from preCICE write output
• tests/unit/micro-manager-config-local-adaptivity-data.json: test config with local_data
• docs/configuration.md: new parameter documented in table and example config updated
• docs/adaptivity.md: new "Local data for adaptivity" section with usage example
• CHANGELOG.md: entry added under [Unreleased]