add ViperJS to the zoo

[MerlijnW70]

Would like to add ViperJS — an embeddable JavaScript
engine in safe Rust, #![forbid(unsafe_code)] crate-wide, with an empty dependency table that CI
checks. RegExp is therefore its own backtracking engine rather than a crate.

The closest comparison targets in the zoo are probably boa and kiesel.

A PR with engines/viperjs/{Dockerfile,Makefile,README.md} and a harness/config.yml entry is
ready in #36 — built and smoke-tested on jsz-rust/amd64. Happy to close it and let you take it
from here instead if you prefer requests to arrive as issues only.

[linusg]

The closest comparison targets in the zoo are probably boa and kiesel.

In which regard?

[MerlijnW70]

Fair question — the sentence compressed three different things into one word, and the obvious
reading of it is wrong. It isn't language: Kiesel is Zig.

Three axes, and the two are closest on different ones.

Design class. From-scratch, no JIT, a bytecode interpreter, and conformance as the objective
rather than throughput. Boa is a stack-based VM, Kiesel a register-based one, ViperJS stack-based;
all three publish a test262 number and are steered by it. That is the axis that makes a zoo row
comparable at all.

Scale and age. Kiesel is the closest row here — 70,969 LOC and 2023- against 61,812
hand-written and 2026-. Comparable order of magnitude, comparable amount of the language still
missing, both still moving.

Embedding posture and host language. Boa is the closest here: Rust, a library first with a CLI
on the side, meant to be embedded in someone else's binary.

And the axis where neither is close, which is the more interesting half: Boa has a dependency tree
and takes regress for RegExp; ViperJS has an empty dependency table checked in CI,
#![forbid(unsafe_code)] crate-wide, and therefore its own backtracking engine. I don't think
there is a close row in the zoo for that, which is part of why I hedged with "probably".

To be plain about what that sentence was: the rows I'd expect a reader to put ViperJS next to, not
a measurement placing it beside them. I have my own test262 figure — 86.41% — but the denominator
is 93,161 runs, each file run in both modes where its flags allow and staging/ excluded, against
your 53,167. It is not comparable with Boa's 96.8% or Kiesel's 93.7% and I would rather it not be
read as if it were. The zoo's harness is the only thing that settles that, which is most of why I
want the row.

Two figures in #36 have gone stale since I opened this, if it's useful — I have pushed the first
and left the second alone. LOC re-measured with the command the README documents is 103,829, of
which 42,017 is four generated Unicode tables, so hand-written is 61,812; the jump since I filed is
normalize and case-folding data rather than engine. And Standard: ES2023 (partial) now
understates it, but whatever convention you apply to the rest of the list is the right one, so I
have not touched that line.