replace bare except: with specific exception types

[Devguru-codes]

Replace all 13 bare except: blocks in LSQ_fitting.py and 1 in triexp_fitting_algorithms.py with specific exception types:

• Parallel fallbacks: except Exception
• curve_fit/minimize failures: except (RuntimeError, ValueError)
• Bayesian fit: except (RuntimeError, ValueError, TypeError)
• Goodness-of-fit: except (IndexError, ValueError)
• NNLS: except Exception

Also fix pre-existing bug: raise(params.message) on line 668 was raising a string (TypeError), not an exception. Changed to raise RuntimeError(params.message).

Fixes #166

[oliverchampion]

It is a while back for me, but I these exceptions were there to make sure the whole program does not fall down if e.g. it encounters a NAN or inf, or some other unexpected insert in one of the voxels. But I agree that the program should break if there is something fundamentally wrong.

Are we sure these are the only exceptions we want to catch to ensure stable behaviour?

[Devguru-codes]

Hello @oliverchampion sir, i apologise for the delay, i was busy finding an internship. I got the time now and reviewed what you said. You were right, i did miss some exceptions. To find everything i wrote a sanity-check script that feeds nan, inf, zeros, huge values, negative signals, and other edge cases into every function. Result was 53/53 bad-data scenarios handled gracefully (no crashes).
I also added some more to be just safe -
a. FloatingPointError on curve_fit blocks (covers NaN/inf when numpy FP error mode is strict)
b. TypeError on goodness_of_fit (len() on a scalar raises TypeError, not IndexError)
c. TypeError on fit_segmented_tri_exp (there's a pre-existing bug on line 454: min(0.1) always raises TypeError since min() needs an iterable. The old bare except: was hiding this.)
I also tested with np.seterr(all='raise') to simulate the strictest floating-point mode. Everything degrades gracefully. Thank you.