Fix startup performance

[jonathanolson]

It's very slow, sometimes ~20 seconds on the iPad. This can consist of:

• Delay reading / computing all structures initially, just the ones needed for display.
• More compact way of storing molecule data, or pre-compress

[jonathanolson]

Big delays are in fromSerial2 (CompleteMolecule and the supertype MoleculeStructure), a bit from PubChemAtom reader
Medium cost in StrippedMolecule creation
Low cost in ElementHistogram (and its getHashString), but that may be lowest hanging fruit.

[Denz1994]

It has been noted that start up time isn't an issue and performance issues haven't been reported in the latest dev-test. See here: #109.

[ghost]

As of today (2020-01-23), I would say the startup time is a bit long.

Testing BAM on master gave the following results. (These times were measured from hitting the reload page button to the time the home screen was loaded.)

Device	Time (s)
newish macOS (my MacBook Air)	8.49
newish Win10 (Logan's Lenovo)	15.79
newish iOS (my iPhone SE)	13.7

@Denz1994, if these times are acceptable, feel free to close.

Also, the loading bar is empty for a decent chunk of the loading time, which might make a user think the sim is having trouble loading.

I can also do more testing if necessary.

[ghost]

@Denz1994, I just tested https://phet-dev.colorado.edu/html/build-a-molecule/0.0.0-dev.47/phet/build-a-molecule_en_phet.html and the performance was much faster on all the above platforms. Load time took ~5 seconds on all platforms, and this seems acceptable.

[Denz1994]

Thanks for double-checking on the start up time. Closing this one.