Compiled artifact carries no compute-budget estimate; required commitment is discoverable only by node rejection

[mw-pearson]

Summary

Spending a compiled contract requires committing a computeBudget on the
input. The artifact JSON contains no estimate of what a contract's entry
will cost, and I could not find a documented way to obtain one, so the
working procedure on TN10 is: submit, read the rejection, commit the printed
number.

Environment: silverscript v1-rc1 (c7d17a1) · rustc/cargo 1.98.0 · TN10
node kaspad v2.0.1 · macOS (Apple Silicon).

Current behavior

Measured example (contract: one r0.succinct.verify call with a
222,668-byte seal, plus output-value and output-script introspection):

• Commit 2,501 → rejected: script units exceeded the amount committed in the input: used=25446424, limit=25019999
• Commit 2,543 → rejected: used=25446424, limit=25439999
• Commit 2,544 → accepted (TN10 txid
  3b32c7bdcf22776c78bae81e4fd40fcfedeab994b94800997c60ddffc2cfde50, and
  boundary-tested again at
  e409473c97d470d0b87ac909cb24e923bad707f3d704127a89d6fca2301764a3).

Two measurements that may help scope a fix:

• used was identical (25,446,424) across two different receipts of the
  same byte-length through the same script — the meter appears
  payload-invariant, so a static per-entry estimate looks feasible.
• The compiler already does conservative per-entry sizing for the
  signature-script-limit check
  (compiler_rejects_entry_abis_that_cannot_fit_the_unlocking_stack), so
  machinery for per-entry static analysis exists.

Expected behavior

An estimated (or upper-bound) script-unit cost per entry in the artifact
JSON, or a silverc flag that prints one.

Impact

Tooling ergonomics. Trial-by-rejection works on a testnet; it is a poor fit
for anything unattended.

Acceptance criteria

• A per-entry cost estimate is obtainable from the artifact or the CLI
  without submitting a transaction.