Spoofing-on debug instrumentation (counters + startup log)

[ryanbr]

Three small additions to make the opt-in spoofing path easier to diagnose and A/B against the now-default-off state. Pure observability — no behavioral change to spoofing, ad-blocking, or the GQL beacon flow. No version bump per the no-per-PR convention; ## Unreleased bullet under ### Diagnostics.

What this PR does

(1) Startup log when spoofing is enabled. After the twitchAdSolutions_disableAdSpoofing localStorage handler, an additional if (!DisableAdSpoofing) block fires a clear startup line:

[AD DEBUG] AdSpoofing ENABLED (localStorage opt-in) — firing GQL ad-tracking beacons on every ad-laden poll. Default is OFF as of v68.3.0; opt-in mode for A/B testing. Watch session counters appended to "Spoofed ad completion" / "CSAI ad request detected" lines to compare with a spoof-off session on the same channel.

Source is named distinctly: localStorage opt-in vs modified default (in case someone patched the default in their own copy). Replaces the previous shorter "enabled via localStorage opt-in" log.

(2) Worker session counter on Spoofed ad completion:

- ... pod-complete: yes
+ ... pod-complete: yes [session: 7 ads spoofed]

Counter held on notifyAdComplete.sessionAdsSpoofed (function-property pattern matching the existing loggedNoMatch / loggedBadStatus style — survives across calls within one worker instance).

(3) Main-thread session counter on CSAI ad request detected:

- ... not blockable via m3u8)
+ ... not blockable via m3u8) [session: 4 CSAI requests so far]

Counter (sessionCsaiRequests) declared at module scope, incremented on every edge.ads.twitch.tv request (placed before the existing once-per-type dedup gate, so the value shown is the true running total at the point of first-of-each-type emission). Counts every CSAI request, not just unique types.

What this enables

Two sessions on the same channel — one with twitchAdSolutions_disableAdSpoofing='false' (spoof-on opt-in), one with default-off — become directly comparable on:

• Total ads spoofed across the session
• Total CSAI requests delivered to the session

Whichever is bigger (or smaller) under spoof-on vs spoof-off is the actual empirical answer to "does spoofing affect ad delivery" — which the disable hypothesis (PR #232 / v68.3.0) has been unable to test without this kind of counter. The localStorage opt-in was already there; this just makes the A/B's output legible at a glance.

Scope notes

• All 4 vaft files share the Spoofed ad completion line — same transform on each. (1) and (2) shipped to the testing pair already at master commit 73ee921. The release pair adds (1) + (2) + (3); the testing pair doesn't have the edge.ads.twitch.tv fetch-hook detection, so the CSAI counter is release-only.
• video-swap-new: spoofing is vaft-only; no port needed.
• acorn-clean on both release files.

🤖 Generated with Claude Code