Map improvements including transmission ribbons

[trvrb]

This PR includes a few separate map improvements.

1. Slightly jitter each separate transmission line so that if there are multiple events they appear as a ribbon rather than a narrow line. I think this is a significant improvement. And it has the nice effect where the transmission ribbon thickness properly interacts with visibility adjustments made through date slider or tree zoom.

Here's current Ebola:

Here's an animation of flu spread:

This just adds an additional extend parameter that is propagated down through the map code to the underlying bezier math that pushes the arc out by extend number of pixels. extend for a particular pair of demes starts at 0 and increments with each transmission event.

This was mainly motivated by the ugliness of current regional flu view, but I think made a big difference overall. Compare before:

To after:

2. Improve deme circle sizing. Previously, deme circle area was proportional to the number of tips belonging to this deme. This resulted in large datasets like flu having circles that were too big and small datasets like Ebola in North Kivu having circles that were too small. This PR adds a normalization by the square root of the total number of nodes in the tree. I kept this as square root so that smaller datasets still have slightly smaller circles than larger datasets. Within a single auspice view, we maintain circle area as proportional to tip count (as I believe is most expected behavior).

This mostly addresses #280. Initial deme radii look good. Zooming out too far on a narrow set of regions (like Ebola in West Africa or H7N9 in China) doesn't look great, but it's really not bad.

3. Improve initial map bounds logic. This is a small fix to make initial zoom bounds better fit to the lat/longs that are present. Essentially we just needed to set Leaflet's zoomSnap to something more granular than the default 1.

4. Don't show error for missing lat/longs if deme is labeled "Unknown" or "unknown". This was apparent in https://nextstrain.org/community/inrb-drc/ebola-nord-kivu, where the samples that are labeled as "unknown" really should not have an associated lat/long.

5. Add very slight border to deme circles. I think this looks stronger and more grounded.

Resolves #425.
Resolves #369.

[rneher]

I like it. looks much better IMO!

[trvrb]

I'm going to go ahead and merge this as well.

[rneher]

There is one aspect that looks a little strange:

The circles around the disk have sometimes odd colors. This is from
https://nextstrain.org/flu/seasonal/h3n2/ha/6m?c=gt-HA1_91&p=grid

[trvrb]

Thanks for catching this Richard. I know what this is (stroke color not updating on colorBy change). Edit: Fixed in: cefef4a

[jameshadfield]

Looks great!

[ghost]

cool！