Widen the ending glitch window; decouple jitter duration from it

User reported the progress bar consistently stalling a few segments
short of full -- SUB/WAVE's reported track duration appears to run a bit
longer than the station actually plays before crossfading/cutting to the
next track, so "remaining" rarely if ever dropped below the old 5s
end-window before the subsonic_id changed out from under it. Split the
single STATIC_OVERLAY_SECONDS into separate start (5s, unchanged) and end
(22s) windows to compensate.

Widening the end window naively (scaling GLITCH_REPEATS to cover all 22s
of it) pushed a real render to 481KB, over the 460KB device limit --
each repeat adds several more uniquely-tinted/shifted frames for WebP to
encode, and 22s of continuous jitter is a lot of those. Added
GLITCH_ACTIVE_SECONDS (6s) to cap how much of the window actually gets
jitter bursts, independent of how long show_static itself stays true --
scanlines and corner ticks are cheap, constant per-pixel layers that
still cover the full window on their own.

Verified: payload stays well under budget (max 234KB) across every zoom
level/position combination after the fix.

subwave_nowplaying.star

@@ -91,11 +91,21 @@ COLOR_SCHEMES = {
 }
 DEFAULT_COLOR_SCHEME = "cyberpunk"  # matches the station's own site theme
 
-# How many seconds into a track's start, or before its end, the "static"
-# overlay shows over the album art (a bundled animated GIF of sparse noise
-# specks over a transparent background -- images/static.gif). Only used
-# when the "Static overlay" schema field is enabled.
-STATIC_OVERLAY_SECONDS = 5
+# How many seconds into a track's start, or before its reported end, the
+# glitch overlay shows over the album art. Only used when the "Static
+# overlay" schema field is enabled.
+#
+# These aren't symmetric on purpose: SUB/WAVE's reported `duration` is a
+# track's full metadata length, but in practice the station appears to
+# cut/crossfade to the next track a bit before that duration actually
+# elapses (observed via the progress bar consistently stalling a few
+# segments short of full, never reaching the end) -- so `remaining` was
+# rarely if ever dropping below a short end-window before the subsonic_id
+# changed out from under it. Widening the end window compensates for that
+# observed early cutoff; the start window doesn't have the same problem
+# (elapsed genuinely starts at 0) so it stays short.
+STATIC_OVERLAY_START_SECONDS = 5
+STATIC_OVERLAY_END_SECONDS = 22
 DEFAULT_STATIC_OVERLAY = "on"
 
 # How far in the album art zoom goes, relative to its fully-zoomed-out rest
@@ -180,7 +190,7 @@ def main(config):
     show_static = (
         static_overlay_enabled and
         progress != None and
-        (progress["elapsed"] < STATIC_OVERLAY_SECONDS or progress["remaining"] < STATIC_OVERLAY_SECONDS)
+        (progress["elapsed"] < STATIC_OVERLAY_START_SECONDS or progress["remaining"] < STATIC_OVERLAY_END_SECONDS)
     )
 
     # The show/host and time/temp rows always render, even when empty, so
@@ -384,14 +394,24 @@ GLITCH_STATE_POOL = [
 ]
 
 # render.Sequence doesn't loop on its own (see _glitch_scanlines below), so
-# to repeat the burst-then-pause pattern for the whole show_static window,
-# the repeats have to be spelled out as one long children list up front
-# rather than relying on any runtime looping. Rounds up so the window is
-# always fully covered, even if that runs a little past
-# STATIC_OVERLAY_SECONDS -- better than the jitter cutting out early.
+# to repeat the burst-then-pause pattern, the repeats have to be spelled
+# out as one long children list up front rather than relying on any
+# runtime looping.
+#
+# Deliberately *not* sized against the full show_static window (which can
+# now be up to STATIC_OVERLAY_END_SECONDS = 22s): each repeat adds several
+# more uniquely-tinted/shifted frames for WebP to encode, and scaling that
+# up to cover 22s of jitter pushed a real render past the 460KB device
+# limit (measured 481KB). Scanlines and corner ticks (see _glitch_effect)
+# are cheap, constant per-pixel layers that can cover the full window for
+# free, but the hue-shifted bands don't need to keep jittering the whole
+# time to read as "something's happening" -- GLITCH_ACTIVE_SECONDS caps
+# how much of the window actually gets jitter, independent of how long
+# the window itself is.
+GLITCH_ACTIVE_SECONDS = 6
 _GLITCH_CYCLE_TICKS = GLITCH_STATES_PER_CYCLE * GLITCH_HOLD_TICKS + GLITCH_PAUSE_TICKS
 _GLITCH_TICKS_PER_SECOND = 20  # 1000ms / 50ms-per-tick
-GLITCH_REPEATS = int(STATIC_OVERLAY_SECONDS * _GLITCH_TICKS_PER_SECOND / _GLITCH_CYCLE_TICKS) + 1
+GLITCH_REPEATS = int(GLITCH_ACTIVE_SECONDS * _GLITCH_TICKS_PER_SECOND / _GLITCH_CYCLE_TICKS) + 1
 
 def _glitch_states_for_cycle(cycle_index):
     # Rotate through GLITCH_STATE_POOL instead of slicing (which doesn't