Copasi-basico

This is a bug fix release silencing a misleading error message coming from the steady state task

This release adds access to the results of the Linear Noise Approximation Task. (it requires an `python-copasi > 4.44`). For an example of running the task see the:

* [LNA Task Example](https://basico.readthedocs.io/en/latest/notebooks/LinearNoiseApproximation.html)

Generally, the LNA is computed runing `run_lna` the optional argument `return_results` (default `False`) indicates whether the full result is returned, or only the status of the computation. To retrieve the results after a computation of the LNA the following methods may be used:

* `get_lna_status` returns the status of the last computation
* `get_lna_covariance_matrix` returns the full covariance matrix
* `get_lna_reduced_covariance_matrix` returns the reduced covariance matrix
* `get_lna_reduced_b_matrix` returns the reduced b Matrix
* `get_lna_solution` returns the tuple: `(status, covariance matrix, reduced covariance matrix, reduced b matrix)`

This issue changes the behavior when running time courses, obtaining values only at specified time points. Previously the first value might have been the first value specified or an earlier time point if the model was not currently at the time. This release changes that behavior to set the output start time to the first specified value.

Additionally, the previously private `collect_data` method can now be used to obtain values from the current model state using either display names or cns.

This resolves issue 61.

This is a bugfix release that fixes an issue that occurs if basico is used with the python-copasi package 4.45 and using optimisation constraints.

This release fixes an issue with `load_model_from_string`, where specific SBML models were mistakenly imported as COPASI models.

This release fixes an issue where returning simulation results from `run_time_course` could not be obtained if multiple model elements were named `Time`. If this issue is encountered:

* basico will no longer try to set the index, and instead issue a warning `Duplicate column names in time series consider using use_sbml_id=True, or running ensure_unique_names()`
* the new function `ensure_unique_names()` appends a count to the name ensuring uniqueness

This fixes 60 .