Aardvark.Dom 1.1.8

• • Aardvark.Application (>= 5.6.4)
  • Aardvark.Base (>= 5.3.20 && < 5.4.0)
  • Aardvark.Dom.Core (= 1.1.8)
  • Aardvark.Geometry (>= 5.3.20)
  • Aardvark.Rendering.Text (>= 5.6.4 && < 5.7.0)
  • Aardvark.SceneGraph (>= 5.6.4 && < 5.7.0)
  • FSharp.Core (>= 8.0.100)

fix: MSAA pick decoder could return the wrong scope at silhouettes when scopes of different modes shared an id space. Specifically — at a silhouette where MS resolve averaged a mode-A pickId with the cleared `0` background (e.g. Green `+2` and `0`), the result `+1` would round to a clean integer and look up `scopes.[1]`. If id 1 happened to belong to a *mode B* scope (e.g. Blue, encoded `-1`), the decoder returned Blue's scope but interpreted the pick attachment with mode A's layout — giving a garbage view-position to the wrong scope. Cross-mode silhouettes (mode A vs mode B pickable) had a symmetric flaw.
introduces a `pickModes : Dict<int, bool>` populated at wrap time from `chooseChain.Final` (`true` = mode A, `false` = mode B). The decoder now rejects any candidate whose `sign(pixIdRaw)` doesn't match the registered scope's mode, before the layout-dependent decoding runs.
`readIdAt` now also rejects non-integral reads — anything more than `0.01` from `round(f)` decodes to `0`. MSAA-resolved pickIds are only trustworthy when all 4 samples agreed (→ exact integer); averaged silhouette mixes that happen to land on a clean integer for the *center* pixel are caught instead by the neighbour validator (since adjacent pixels with different coverage produce different averages).
same-id neighbour validation tightened from ≥ 2 to ≥ 3 of 8. Two adjacent silhouette pixels with similar coverage can produce the same near-integer averaged id; three in a 3×3 essentially never do.
`pickModes` is cleaned up in `releaseId` alongside the other id-state dicts.