Ref T13279. Since the use cases that have made it upstream are all for relatively complex charts (e.g., requiring aggregation and composition of multiple data series in nontrivial ways) I'm currently looking at an overall approach like this:

• At least for now, Charts provides a low-level internal-only API for composing charts from raw datasets.
• This is exposed to users through pre-built SearchEngine-like interfaces that provide a small number of more manageable controls (show chart from date X to date Y, show projects A, B, C), but not the full set of composition features (compose(scale(2), cos()) and such).
• Eventually, we may put more UI on the raw chart composition stuff and let you build your own fully custom charts by gluing together datasets and functions.
• Or we may add this stuff in piecemeal to the higher-level UI as tools like "add goal line" or "add trend line" or whatever.

This will let the low-level API mature/evolve a bit before users get hold of it directly, if they ever do. Most requests today are likely satisfiable with a small number of chart engines plus raw API data access, so maybe UI access to flexible charting is far away.

Step toward this by adding a "Reports" section to projects. For now, this just renders a basic burnup for the current project. Followups will add an "Engine" layer above this and make the chart it produces more useful.