Fix review findings and add regression tests

[sergioald]

This PR addresses six issues identified during an independent review of LDSFL_Meander.

Fixes:

• Correct direct bed-shear-stress to Shields-stress conversion.
• Preserve previous simulation outputs by isolating repeated runs.
• Preserve dimensional output scaling during GUI continuation.
• Select batch cases using the actual Parameter.csv IDs.
• Correct stop_mode="all" handling for inactive limits.
• Avoid copying full sinuosity histories for moving-window diagnostics.

Additional changes:

• Added regression tests for the corrected behaviour.
• Updated user documentation and changelog.

Validation:

• 175 tests passed
• 1 environment-dependent test skipped
• 1 pre-existing expected failure remains for the known Numba SL=1 discrepancy