Use a color encoding on the heatmap that is easier to decipher

[jeremiak]

Relevant feedback from Google doc:

Add median download indicator to heatmap

Comment from Slack:

I’m still not sure that the heatmap coloring is noticeable enough for me. I can’t quite tell it apart. I would probably get the mean of all the speeds within a bucket (across all days) and map that to a neutral color, like a white. I would then pick the max and min and have those be the brighest colors, while closer to the mean is neutral, so white or whatever the background would be. I would also use a logarithmic scale to really highlight the differences, since I imagine they would be quite small. so [min, mean, max] -> [min tone, neutral invisible color, max tone] - whatever the colors make sense in that setup.

[jeremiak]

Using @iros's feedback from Slack ^^ to highlight the differences in the distribution of the speed tests.

Heatmap viz methodology

1. Take all of a location's bucketed data and combine some of the buckets so that they have more human meanings, for example having a break at 2mbps and 10mbps since those have real world consequences. Recompute the number of samples in the new buckets.
2. Use all of the dl_frac values to find the mean, min, and max. In this way, we are measuring which buckets have more or less than the average number of tests in them.
3. Encode the mean value to a gray, while encoding the min to green and the max to pink

Screenshot:

Because of the way we're doing the Y axis on the heatmap chart, however, I'm unsure how we want to add a median speed indicator.