README: add diagrams for noop and trailing-stop behaviors

[mellonis]

Motivation

Two PostMachine conventions are mentioned in prose but not visually illustrated. Both are subtle enough that a small Mermaid diagram alongside an example would shorten the path from "I read the docs" to "I understand what the graph looks like."

(1) noop — the fall-through pass-through

noop is the "do nothing, advance to the next instruction" command. Visually it's a state with a single [*] → [K]/[S] self-targeting transition to the next instruction's State. Useful when teaching:

• The pattern noop(20) as an unconditional jump from any instruction to 20.
• How grouping interacts with noop (grouped noop becomes a no-op step in the chain).
• Why noop is sometimes preferred over right(20) etc. — when you don't want to mutate the tape.

Suggested addition: short <details>-collapsed Mermaid block in the Commands or Quick start section showing a { 10: noop(30), 30: mark, 40: stop } machine. Pair with an examples.spec.ts test pinning the salient shape (one outgoing [*] arrow, no tape mutation).

(2) Trailing stop doesn't produce its own state node

Currently the Quick start prose hints at this: "The 40: stop instruction is elided — stop halts the machine, so the transition from 30: mark flows straight to halt rather than through an intermediate state."

But the visual consequence isn't pinned with its own diagram. A reader who tries { 10: mark, 20: stop } and inspects the emit might be surprised to see only ONE non-halt state (s10), not two.

Suggested addition: a tiny Mermaid block alongside the prose showing:

flowchart TD
  s0(((halt)))
  s1["10"]
  idle([idle])
  idle -. enter .-> s1
  s1 -- "[*] → ['*']/[S]" --> s0

Loading

— and an explicit "notice: no s20 for the stop; the engine routes directly to s0" line.

Scope

• Single-package change (@post-machine-js/machine README).
• Two added <details> blocks + two test assertions in examples.spec.ts.
• No code change. Strictly documentation.

Cross-refs

• Surfaced during the #85 drop-hopper refactor — while reviewing structural-summary diagrams it became clear other PostMachine conventions deserve similar visual treatment.
• Companion to the alpha.2 PR's structural-summary diagram addition (PR chore(release): 7.0.0-alpha.2 — engine alpha.2 adoption (closes #82, #83) #84).

[mellonis]

Re: pm.stateAt for trailing-stop indices

Probe confirms: today pm.stateAt({ instructionIndex: 20 }) for { 10: mark, 20: stop } returns haltState (the singleton). So the convention is asymmetric in a useful way:

• Graph emit: trailing-stop indices are elided — no s20 node appears. Halt-bound transitions route directly to the s0(((halt))) singleton.
• Lookup API: pm.stateAt(<trailing-stop-index>) resolves to haltState rather than throwing.

The README diagram should highlight this — readers seeing "no s20 in the diagram" might assume pm.stateAt(20) would throw, when it actually returns the canonical halt sentinel. Worth one prose line alongside the diagram.

[mellonis]

Shipped in v7.0.0-alpha.3 via #90. Closing for v7 milestone bookkeeping. (Release PR #92 still references this issue for historical trace.)