orbit lab: drop the record_history opt-out once EigenScript#827 lands, and re-point planted fault 1

[InauguralPhysicist]

Follow-up to the rung-1 UI ladder (#20), specifically the memory-gated
bifurcation slice (#23).

What changes upstream

EigenScript#827 — temporal assignment history is unbounded and arms on dead
code — is fixed by InauguralSystems/EigenScript#829. Two things we depend on
here go away:

1. The history is now bounded by program TEXT, not by runtime. Entries a
   backward query can never reach are pruned at append time, so a render loop
   that reassigns the same names every frame keeps a handful of entries and
   pins a handful of values instead of growing without bound. Measured
   upstream: a live prev of loop at 1.6M iterations went 203,008 kB → 2,944 kB
   (the no-temporal-query floor), flat in iteration count.
2. Arming is per-name, not whole-program. physics.eigs' single prev of
   in the never-called frame_velocity demo no longer arms recording for the
   whole program — it arms only the name that query mentions. The dead-code
   repro went 53,120 kB → 2,944 kB.

Semantics are unchanged, so nothing about our temporal reads shifts.

What that breaks here

orbit.eigs:707 calls record_history of 0 for the lifetime of a lab window
session (restored at :729), with a long comment at :691 explaining why. That
workaround becomes unnecessary — but more importantly, planted fault 1 in
tests/test_bif_mem.sh:205 will stop failing:

T1=$(plant hist "    local hist_was is record_history of 0" \
                "    local hist_was is record_history of 1")

It flips the opt-out on and requires the gate to reject the resulting build. Once
#827 is fixed, turning history back on no longer produces the 3.9 MB/frame
growth, so the fault is caught by nothing and the gate's self-test goes green for
the wrong reason — a planted fault that can no longer be planted is a gate that
has quietly stopped discriminating.

The work

• Bump the EigenScript pin to the release carrying EigenScript#829.
• Remove the record_history of 0 / record_history of hist_was pair and the
  :691 comment block from _run.
• Replace planted fault 1 with one that still bites. The gate's job is to
  catch per-frame retention in the bifurcation view; it needs a fault that
  actually reintroduces unbounded growth on the fixed runtime — e.g. removing
  the chart's chart_trim / series cap, or the build-once sweep guard's
  companion. Do not simply delete fault 1: the gate would then be validated by
  one fewer fault than it was designed around.
• Verify the memory gate still passes on a clean build and still fails on
  every remaining planted fault, per the gate's own self-test contract.

Blocked on EigenScript#829 landing and a release cut. Not urgent — the current
workaround is correct and costs nothing; the risk is only that the gate silently
weakens if the pin moves before this is done.