Small speedup improvement on the inspection logic in Python

[martinRenou]

In addition to #309 to get through improving the extension performance #307

---

This PR adds more checks for simple primitive types before doing the "heavy load" for more complex types.

In this situation:

It goes from:
803 ms ± 19.1 ms per loop (mean ± std. dev. of 7 runs, 1 loop each) for _jupyterlab_variableinspector_dict_list
to
265 ms ± 14 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

Note that this means without #309 and this PR we were adding almost 1 second delay for each cell execution when the variable inspector UI was not open. Even though I'm not testing with so many variables and my arrays aren't too big.

This is a first PR of hopefully more performance improvements

[github-actions]

👈 Launch a Binder on branch martinRenou/jupyterlab-variableInspector/small_speedup

[martinRenou]

cc. @trungleduc @achhina

[krassowski]

I guess a lot of the penalty is in stringification:

if str(obj)[0] == "<":
    return False
if  v in ['__np', '__pd', '__pyspark', '__tf', '__K', '__torch', '__ipywidgets', '__xr']:
    return obj is not None
if str(obj).startswith("_Feature"):
    # removes tf/keras objects
    return False

If object takes long to convert to string we are wasting a long time doing it twice just to peek on whether it starts with < or _Feature. We can probably save some milliseconds by moving the str() call to end or ideally getting rid of it altogether (I think for tf/keras we could look at str(type(obj)) to avoid importing the thing)

[krassowski]

Looks good to me, again let's merge the main in to make sure CI passes

[martinRenou]

Thanks for your suggestion, I will have a try