Use smaller data set for profiling

[MarionBWeinzierl]

The current profiling functionality relies on a rather larger data set (1.8GB) and then uses just a manageable slice. See also https://github.com/ImperialCollegeLondon/pyrealm/blob/ad32a8eff8ea42ded2a6cefc0d33880e8560332b/tests/pmodel/test_profiling.py#L27 .

This needs some looking into.

Options to consider:

• Keep the large data set and define a minimum slice that still gives reasonable results but reduces runtime.
• Find/build a smaller data set.
• Create a smaller synthetic data set (which might not even need be populated with realistic values).

[MarionBWeinzierl]

In the long term, it would be more elegant to generate the data set. But it's good for now.