FAT workstation convergence: architecture freeze and canonical shell staging

[masarray]

Purpose

Stage the professional Engineering-integrated FAT convergence without destabilizing PR #290's current known-good baseline.

This PR remains intentionally stacked on feat/native-fat-workspace (PR #290), not main. It is the protected staging lane for the architecture cleanup and field gates.

Current implementation state

• canonical seven-destination Engineering shell with permanent FAT destination
• FAT bootstrap reuses the already-parsed Engineering SCL/static DataSet authority
• Engineering remains the single parsed SCL/live-value authority; FAT is an evidence/session projection
• Start FAT preflight now self-heals only a proven generated scl-manual-* shadow when an authoritative Static DataSet row resolves to the exact same live IEC 61850 reference
• restored/manual TEST preference, FAT disposition and evidence are preserved; only duplicate shared-workspace ownership is retired
• truly ambiguous non-manual duplicates still fail closed with the live IEC reference included in the error
• same-source continuation preserves saved FAT progress without replacing the fresh Engineering model
• source staging reuses an already-verified content-addressed staged SCL copy instead of rewriting identical bytes on every FAT-tab entry
• no second navigation owner / launcher parity shim is reintroduced

Protected behavior

No FAT change in this staging branch is allowed to introduce Hybrid/cyclic MMS polling for static DataSet FAT, restart the shared Engineering acquisition session, silently duplicate static DataSet members, weaken evidence/source integrity, or bypass strict duplicate-live-reference preflight.

CI gate — GREEN on current head

Head: f901bb31c9a59ee0796afc86a8679b8dd6da1f22

• Build ARSAS #2578: success
• full application regression step: success
• Validate IO List Testing #1131: success
• Validate SV evidence bundles #1580: success
• win-x64 portable single-EXE publish: success
• portable EXE smoke test: success
• field regression covers the exact restored scl-manual-* live-reference shadow class
• regression also proves a true non-manual duplicate remains blocked

Remaining gate — FIELD

Use the exact ARSAS-win-x64-portable-single-exe artifact from Build ARSAS #2578 (ARSAS-1.6.35-win-x64-portable.exe) and verify:

• reproduce the same IED/SCL used by the failed 39997457 field candidate
• Start FAT must no longer reject the scl-de... + scl-manual-... pair when both resolve to the same proven live leaf
• a genuinely ambiguous duplicate mapping must still be rejected
• FAT tab first/repeated entry latency remains acceptable
• Engineering -> FAT does not reconnect, enable Hybrid or start cyclic MMS polling
• static DataSet row count, live value, quality, relay timestamp, Value 1 / Value 2 capture and evidence remain correct
• Report Preview/export remains unchanged and correct

Keep this PR draft and do not merge into PR #290 until this exact field candidate passes the field gate.