Fix problems reading .mat files

[lfarv]

This fixes #340.

• Workaround a bug (?) in scipy.io.loadmat where empty strings from .mat files disappear,
• Fix problems with particle objects in .mat, .m and .repr files

[simoneliuzzo]

save/load of a mat file from python does not give errors

[simoneliuzzo]

sorry I have to take it back. I tried an other file.

a file saved in python has the RF cavities with DriftPass, and gives this error:

AttributeError: Error in element 136: PassMethod DriftPass is not compatible with Class RFCavity.
{'FamName': 'RFC', 'Length': 1e-12, 'PassMethod': 'DriftPass', 'Voltage': 9000000.0, 'Frequency': 352371625.9975046, 'HarmNumber': 979.0, 'Energy': 6000000000.0, 'TimeLag': 0.0}

I think the problem rises when saving a lattice with radiation off (that turns also the cavity off)

[lfarv]

@simoneliuzzo : this is a different problem. The "coherence" test when reading a file is too strict. We'll change that: I agree that DriftPass may be use for any class. For the moment, you can skip the test by adding "check=False" to the arguments of the load function (load_lattice or load_mat).

[lfarv]

@simoneliuzzo: side remark: you should NEVER use lengths like 1.0E-12.

• for cavities, it's not dramatic: it just adds a half-long drift (0.5E-12) on each side of the the (thin) cavity,
• for multipoles it's much worse: apart from being about 20 times slower than a thin multipole (slices, drifts and kicks) thus increasing the numerical errors, it may be wrong because some computations like sin(x)/x may be inaccurate for small arguments.

[lfarv]

The coherence test is relaxed so that 'DriftPass' is allowed for any Element class.

[simoneliuzzo]

it works for me!