fix(runtime): resolve the na helper to NaN in Series.from() (UDT = na defaults)

[alexgrover]

Problem

Running a Pine v6 script with a user-defined type whose fields default to na (e.g. LuxAlgo's RSI Divergence: Out-of-Sample Optimizer) throws on the first bar:

Cannot call 'array.unshift' with argument 'value'='[object Object]'.
An argument of 'literal object' type was used but a 'float' is expected.

Root cause

• type SimState declares float rsi = na. The transpiler emits the default as ['float', na], where na at runtime is the NAHelper singleton.
• Type.new resolves defaults via Series.from(default).get(0). Series.from() only unwrapped helpers whose __value is a Series (time, time_close); NAHelper.__value is a scalar NaN, so the helper fell through and was wrapped as an opaque object. The field ended up holding the helper itself.
• na(obj.x) masked the problem (NAHelper.any special-cases the helper), but nz(obj.x, 50) only checks null/undefined/NaN, so it returned the helper unchanged. On bar 0 the script hits exactly sim.rsi := nz(sim.rsi, 50) and then sim.rsi_hist.unshift(sim.rsi), which rejects the object.

Fix

Series.from() now resolves any __value helper whose value is a scalar (the na helper) to that scalar. This is the common unwrap point, so it fixes UDT = na defaults, nz(na), math.* on such fields, and any other consumer in one place. Duck-typed on __value to avoid a Series → Core circular import.

Tests

• New regression test in tests/core/udt-field-defaults.test.ts (explicit \= na` defaults resolve to a real na (NaN), not the na helper object`). Fails with the exact error above before the fix; passes after.
• Full suite: 175 files / 1878 tests pass.
• The original reporting script now runs end-to-end on the Mock provider (690-bar RSI plot with zero NaNs, divergence drawings emitted, dashboard table populated).

---

Note

Medium Risk
Central conversion in Series.from() affects any code path that unwraps __value helpers, though the change is narrow and well-covered by tests.

Overview
Fixes UDT fields declared with = na (e.g. float x = na) storing the runtime NAHelper object instead of NaN, which broke nz(), arithmetic, and typed array operations that expect scalars.

Series.from() now unwraps duck-typed helpers with __value: it still returns backing Series instances for dual-use helpers like time, and for scalar __value (including na’s NaN) wraps that value in a one-element Series instead of treating the whole helper as an opaque object.

Adds a v6 UDT regression test covering nz(obj.x, …), na(obj.x), and obj.x + 1 so defaults behave like real na end-to-end.

^{Reviewed by Cursor Bugbot for commit a606e27. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_325eab94-ca17-4a79-ba82-29b87330c6a2)