Replace convert_objects

[philipstarkey]

Original report (archived issue) by Jan Werkmann (Bitbucket: PhyNerd, GitHub: PhyNerd).

---

There is a deprecation warning that lyse currently throws the first time the dataframe is requested by a analysis routine.

#!python

WARNING lyse: /Users/janwerkmann/labscript_suite/lyse/__main__.py:174: FutureWarning: convert_objects is deprecated.  Use the data-type specific converters pd.to_datetime, pd.to_timedelta and pd.to_numeric.
  convert_dates=False, convert_numeric=False, convert_timedeltas=False)

We should maybe replace it?

[philipstarkey]

Original comment by Philip Starkey (Bitbucket: pstarkey, GitHub: philipstarkey).

---

Hmm, this looks messy. There's a discussion here about the deprecation. There is now a infer_objects method in pandas, but I don't think that directly replicates what we currently do (I think it will probably convert dates and timedeltas - see the source code for that vs convert_objects).

There is also a _convert method we could use, but that's a little dodgy since it's not part of the exposed API.

[philipstarkey]

Duplicate of #52 and also resolved in 9fac3f5 (BitBucket PR 70)