tests: fix the "filter flake" and the view_toggle cascade — suite is 187/0

Neither was a flake; both were test bugs that happened to be timing/binary
dependent.

1. filter: the scenario hardcoded the glob 'sub_1*', which matches NOTHING in a
   binary whose code never reaches 0x1xxx — echo's functions are sub_2xxx..
   sub_7xxx, so it failed deterministically, every run. Derive the glob from real
   names instead (first sub_ prefix present) and assert the row count equals the
   expected match count, which is stronger than the old 0 < n < total. A
   self-check asserts the derived glob actually matches something, so this can't
   silently rot again on another binary.

2. view_toggle: "tab switches to disassembly" asserted after a fixed pause(0.1),
   but decomp -> listing runs through _toggle_to_listing, a BACKGROUND WORKER —
   _active only flips once the listing model has loaded. The pause held in short
   runs and lost the race in a full one, which is why this looked like collateral
   from the filter failure. Wait for the state instead. Two related hardenings:
   focus the decomp pane before Tab (elsewhere Tab is focus-next, silently
   leaving us in the decompiler), and wait for the listing cursor to carry an ea
   before the F5 check (F5 legitimately no-ops on an unaddressed row).

Full suite now 187 passed / 0 failed, the first fully green run — previously
174-182 with 2-6 "known flakes" whose count drifted with machine load.

0 files changed, 0 insertions, 0 deletions