List of performance improvements to be implemented

[jameshadfield]

This issue summarises the available performance improvements that could be made to auspice. These can involve algorithmic improvements and reducing the bundle size(s). Each one would probably be a separate PR.

Code/bundle reduction

• Don't load the Map component for datasets that don't need it (this is easy with code splitting but results in a blank pane for ~0.5s for datasets that show the map).
• Remove font-awesome. We only it in the mobile close/open sidebar icons and can use SVG paths (as done in the static site within nextstrain.org)
• Remove bootstrap CSS -- I believe we are only using this on the status page and the splash page - doing this with Flexbox would save a large CSS fetch.
• Remove lodash -- this is 25kb of the initial bundle (160kb) and we only use a dozen easy-to-reimplement functions.
• Investigate tree-shaking to remove unused code from the bundle. (nextstrain/nextstrain.org@699e258 makes me think we're not doing this).
• Investigate browser specific bundles. Currently the initial 160kb bundle has (at least) 28kb of polyfills which shouldn't be needed for modern browsers. Additionally, native code should run faster than polyfilled code.

Algorithm improvements

• Color averaging performance -- Color averaging performance #289