Slit scan guess

[jackbdoughty]

Slit scan guess was sometimes strugglnig to converge due to initial guesses outside of the data, this causes all differentials to go to zero for some parameters and hence optimizer cannot work.

Simplified heuristic guessing function that keeps all initial parameters within the bounds of the data seemed to work well to give the optimizer a good chance of converging to a decent fit:

            background = np.min(y)
            inflection0 = np.min(x) + (1/3) * (np.max(x) - np.min(x))
            inflections_diff = (1/3) * (np.max(x) - np.min(x))
            gradient = 2 * (np.max(y) - np.min(y)) / (np.max(x) - np.min(x))
            height_above_inflection1 = (np.max(y) - np.min(y)) / 5.0

Needs unit tests etc adding

[jackbdoughty]

See https://github.com/ISISComputingGroup/ibex_bluesky_core/pull/76