fix: add zero photon charge and fix their cutBased id

[ikrommyd]

This adds a zero Photon charge like we do with mass so that operations like diphoton invariant mass (photon1 + photon2).mass don't run into

AttributeError: no field named 'charge'

This error occurred while calling

    numpy.add.__call__(
        <PhotonArray-typetracer [...] type='## * Photon[seediEtaOriX: int8[...'>
        <PhotonArray-typetracer [...] type='## * Photon[seediEtaOriX: int8[...'>
    )

Also fixes the cutBased ID of photons since it was changed in later NanoAOD versions and it is not caught by tests.

[ikrommyd]

Will add a test soon!

[ikrommyd]

I think this is also worth backporting to 0.7

[ikrommyd]

@nsmith- I think it's no longer a bitmap and the documentation is just very bad.
I'm getting zero tight photons within two root files with 1150994 events and 1768508 photons. This isn't expected at all.

In [53]: ak.sum(events.Photon.cutBased & (1 << 2) != 0).compute()
Out[53]: 0

In [54]: ak.sum(events.Photon.isTight).compute()
Out[54]: 0

This is also the distribution of the cutbased id values in those files:

and this is the distribution in our nano_dy.root sample:

The documentation for nanoaod v10 says:

And for v12 it says:

and in v13 they dropped the second cutbased id. I think it was a bitmap when it was documented in the form of "2^(0 : loose ....)".

[ikrommyd]

I don't know if this is the best way to initialize the cutbased attributes in the Photon class. Please let me know about requested changes. The downside with my implementation is that the LOOSE, MEDIUM, TIGHT attributes are not available before a cutbased id method is called. That's only a problem in eager mode though as they are not accessible in dask mode anyways. Doing events.Electron.TIGHT fails with TypeError: meta must be an instance of an Awkward Array, not <class 'int'>.

Another thing I thought of is to start with

    FAIL = 0
    "cutBased selection minimum value"
    VETO = 1
    "cutBased selection minimum value"
    LOOSE = 2
    "cutBased selection minimum value"
    MEDIUM = 3
    "cutBased selection minimum value"
    TIGHT = 4
    "cutBased selection minimum value"

by default and if cutBased is not in fields, then we switch to bitmasking on the spot without changing those attributes. No one is gonna use that old of a NanoAOD version anymore anyways.

edit: I actually did it like that in my commit below.