Log linearized curve overlays

[goodboy]

The cross-symbol y-alignment you always wanted..

Yah it's for #420 🏄🏼 now swapped with #461.

---

wUt dIs?!

• overlays multiple time series using a "log-linearized returns transform", meaning all % returns (moves) on any symbol's chart are the same (from the reference point) as any other overlapped with it.
  • keeps support for multiple Curve per PlotItem y-range sorting
  • there are edge cases we handle:
  • any shorter series is aligned according to its intersect with the longer series it overlays with (in the case of > 2 this series is normally the one measured to have the most dispersion).
  • gaps are handled as though the pre-gap value is the value used as the ref-y-value
• reimplements all auto-y-ranging as part of a new ChartView.interact_graphics_cycle() which does the more efficient pipelining of Viz related graphics update calls per mouse event.
  • use this inside the _display loop as well.
• completely drops all the Qt.Signal stuff for interaction which results in more reliable, easier to understand, and generally faster interaction handling code 😂

---

Solved ToDo bugs:

• [x] we need to add back y-ranging on the vlm chart to make sure the RHS label is always in view (done in 0d45495)

• [x] stupid gap on LHS of vlm fsp chart which still seems to be due to some kinda PlotItem.layout related setting? LUL, fixed in d4262c9

• [x] get the immediate y-range incremental update on L1 working again?

  • not sure how this broke yet again, pretty sure it was just fine a a while back in
  • fixed and improved as of fc0a793

• [x] rewrite commit 325fe3cf which is duplicate of 325fe3cf from #453 after cherry picking it there.

---

Final refinements

• [x] for a single fqsn using array-indexing the x-datetime labels for the cursor still aren't working.. (fixed by 269f659)
• [x] bleh, still an issue with 3+ feeds on slow chart and overlays, eg.
  • HeH, figured it out -> we need to rescale any already scaled minors whenever the major (dispersion) needs to be rescaled (since any previously scaled minors are scaled to the previous major's range 🤦🏼 ) (fixed in 1be5ae4)
• [x] finish formalizing the OverlayT type and API and use lru caching on range (inverse) transform outputs
• [x] support the following 4 viewlist "modes" (finalized in 5e729dd)
  • [x] only one feed
  • [x] ymxmn for each feed
  • [x] pin to dispersion major (1be5ae4)
  • [x] pin from first datum in view (rework in 16eda1e)

---

FINAL FINAL bug fixes and tweaks:

• [x] fix upsampling (re-)triggering when using the 'r' to reset chart hotkey in 1d649e5

• [ ] there's still an issue with de-synced 1m last bars?

• [ ] lol, the alignment to "sigma major" default can definitely still cause a terrible config as per this example of bitcoin overlayed with a deribit call:

  • the issue is that the call option is clearly causing the bitcoin move to be mega squashed in the y-range 😂
  • it might make sense in these cases to instead use a "leverage and delta adjusted" scalar for the opt contract so that you can see a "some what linear" translation of the deriv price vs. the underlying..
  • yet another issue seems to be that the "pin to dispersion major" curve idea doesn't work well when detecting intersects and the sigma major is the shorter curve?

https://user-images.githubusercontent.com/291685/223162334-6d881f36-446e-41f0-a6ea-389f9405ae32.mp4

[goodboy]

Weird, CI breaking on Struct not having a default factor someting somethin?

Not sure when that started?

(fixed with #457)

[goodboy]

tweaked base branch to that from #471 for now just until we either drop those commits from this history or land that branch (which is slightly more critical 😂 )

[goodboy]

Outstanding bullets added to #461.

Think this is ready to land 🏄🏼