Atomica

- Improve error message if calibration file contains duplicate entries

- Transfer parameters no longer raise an error if specified in 'Duration' units
- Transfer parameters in rate units are no longer limited to a maximum value of 1

*Backwards-compatibility notes*

- Transfers in rate units with a databook value greater than 1 were internally limited to a value of 1 previously. Models with such transfers will produce different results. This is expected to be uncommon, as most models have transfer parameters with values less than 1.

- Some numerical errors in `model.py` (particularly relating to errors/warnings in parameter functions) are now caught and printed with more informative error messages.

- Change the table parsing routine again to resolve further edge cases, restore removal of leading and trailing spaces from cells in the framework, and improve performance. The original `None` behaviour has consequently been restored (undoing the change in 1.26.2) although it is still recommended that `pandas.isna()` is used instead of checking for `None`.

- Switch to `sc.gitinfo` from Sciris. The git commit hash recorded in Atomica objects will now only contain the first 7 characters. Code that uses `at.fast_gitinfo` should use `sc.gitinfo` instead.
- Improve robustness of the table parsing routine used by `at.Framework`. In some cases, the data type of empty cells is now `NaN` rather than `None`. This affects any code that either checks for the contents being `None` or which relies on `None` being treated as `False` in conditional statements e.g., `if <contents>:`. Affected code should instead use `pandas.isna()` which handles `None`, `NaN`, and `pd.NA`.

- Improve numerical robustness of `SpendingPackageAdjustment` under certain edge cases
- Fix bug in cumulative plot labelling that could result in the axis label containing more than one 'Cumulative' prefix