Check --no-chronicle results in both eras; write ASM only when it reads back exactly

[codenlighten]

Interpreter tests in both eras

differential() ran under the current (Chronicle) consensus flags only, including for scripts optimized with --no-chronicle, which are meant to hold under the earlier rules as well. The OP_VERIF dead-block bug fixed in 0.1.1 changed behaviour only before Chronicle, so those tests could not catch it.

• differential(a, b, { flags }) now also takes a list of flag sets. It stops at the first era that disagrees and returns that flag set with the counterexample.
• eraFlags({ chronicle }) in src/verify.js: [current], or [current, current without SCRIPT_UTXO_AFTER_CHRONICLE and SCRIPT_ENABLE_CHRONICLE]. Both bits have to be cleared: either one enables Chronicle in the interpreter.
• optimize(…, { chronicle: false }) runs its interpreter tests in both eras. report.verification.differential.eras is 2.

Exact ASM output

--asm -o wrote <tail N bytes> for the data after a top-level OP_RETURN (noted in #1) and OP_PUSHDATA1:<hex> for pushes of 76 bytes or more. Neither reads back, so the written file no longer described the optimized script.

• Output ASM now uses the form @smartledger/bsv reads: data as bare hex, and the OP_RETURN data as ASM.
• exactAsm(script) (exported) returns ASM only if it reads back as the same bytes. Otherwise it throws and names the reason:
  • a non-minimal push (optimized code is always minimal, but OP_0 OP_IF blocks and OP_RETURN data are kept verbatim);
  • a truncated push;
  • an unnamed opcode (0xba–0xfc, see below);
  • data pushes alone, which scriptmin reads back as hex.
• The CLI (optimize and compile) exits 2 with that message instead of writing a different script. Hex and --binary output are unchanged.
• toAsm for display (--explain, --json rewrites) is unchanged.

Tests

• New unit test:
  • round-trip through exactAsm for P2PKH with OP_RETURN data, a 300-byte envelope, and OP_RETURN OP_INVALIDOPCODE 0;
  • each refusal reason;
  • CLI --asm -o writes the OP_RETURN data as ASM and exits 2 for a verbatim OP_PUSHDATA1;
  • eraFlags, and a hand-made wrong optimization that passes under Chronicle flags but is caught in both-era mode.
• New fuzz test: 400 scripts with OP_0 OP_IF blocks holding OP_VERIF/OP_VERNOTIF inside other conditionals, optimized with chronicle: false and checked in both eras. With the 0.1.1 OP_VERIF fix reverted, it fails.
• npm test: 28/28, 34 s locally.

Found along the way: @smartledger/bsv 9.11.0

Script#toASM prints an unnamed opcode (0xba–0xfc, 67 of them) as bare hex, e.g. 51ba → OP_1 ba, and fromASM reads that back as a one-byte data push (5101ba). toString prints 0xba, which fromString rejects. Not changed here; exactAsm refuses such scripts.

[codenlighten]

Reviewed. Both changes checked, and the tests were break-tested:

• Eras: the claim that clearing either Chronicle bit alone leaves Chronicle on matches Interpreter#isAfterChronicle in @smartledger/bsv 9.11.0. With eraFlags reduced to the current era only, a test fails.
• Exact ASM: with the round-trip check in exactAsm removed, a test fails.

Added d033c19, two small fixes:

• dropped the parse import the CLI no longer uses (the only new lint finding in this PR)
• moved differential()'s comment back above it, noting that flags may be a list

The unnamed-opcode ASM round trip reported here is confirmed in @smartledger/bsv and is being fixed there.