nx_zstd_seqtab_gate.nx

nx_zstd_seqtab_gate.nx -- proves the predefined sequence distributions. These are transcribed spec data, so the gate's job is to catch a TYPO. T1 checks the totals directly: 64, 64, 32. A single mistyped digit almost always breaks the sum, which makes this a real transcription check rather than a restatement of the source. T2 checks that all three actually BUILD. nx_fse_build_dtable refuses unless the counts tile the table exactly and every cell receives a symbol, so a successful build is independent evidence on top of the sum. T3 pins that a -1 entry is a REAL low-probability symbol, not a terminator: it must appear in the built table and it must sit at the TOP, which is where the spread algorithm places low-probability symbols. Dropping those entries is the obvious misreading -- they look like end markers. T4 spot-checks distinctive values that a shifted transcription would move: the literal-length 3 at index 25 sitting alone in a run of 2s, and the offset table's 2,2,2 at indices 6-8 surrounded by 1s. license_tier: ORIGINAL