feat(dsl): composition operators, durations and one_of (p8-s2)

[JArmandoAnaya]

A composition contributes exactly one thing to the IR: the start trigger on
a phase's Event. No new runtime concept, no new IR node, no scheduler change —
which is what keeps the two frontends sharing one runtime rather than two that
resemble each other. ADR-0031 records it.

A concrete duration is arithmetic, done at load time

The storyboard starts at t = 0 and every duration that lowers is a constant, so
a phase's absolute start time is the sum of the durations before it. Lowering
tracks, for each point in the do directive, what load time knows:

• an absolute time, when every preceding phase had a concrete duration, and
• a set of groups that must have completed, for phases whose end only the
  runtime knows.

A trigger ANDs whichever of the two it has. A phase starting at t = 0 with
nothing to wait for gets no trigger, because t >= 0 is a tautology and
§7.6.1.1 already says a trigger-less element starts with its parent.

Two obvious alternatives were probed against the engine and both fail, which
is why the arithmetic route is the one taken:

• A stop trigger only cuts a phase short. An Act completes as soon as its
  groups do, so a step-shaped action ends the phase immediately whatever the
  stop trigger says — the duration would be silently ignored for exactly the
  scenarios easiest to write.
• A start trigger delayed on the predecessor's runningState never fires.
  The predecessor is no longer running by the time the delayed lookup happens.

A parallel join is an AND the trigger model already has

A ConditionGroup is a conjunction, so "every member has finished" is one group
with one condition per member. Members ending at a known time collapse into a
single SimulationTimeCondition at the latest of them; members ending when
their actions do each contribute their group's completion — so the two mix for
free, and the common case reduces to one comparison.

§7.6.2.1.4's default overlap: start is exactly what absent start triggers give.
The other seven overlap kinds and start_to_start/end_to_end are reported.

one_of picks by label, defaulting to the first

§7.6.2.1.3's latitude is real and an executor has to pick. Picking at random
would put a hidden input in the run — and the engine has no seed machinery at
all, so such a choice could not even be made reproducible. The alternative is
therefore an input: LowerOptions::alternative, fed by scena-run --select.
Alternatives not chosen never reach the IR, so the engine stays free of a
concept the XML side has no counterpart for.

wait elapsed(d) lowers to nothing but the clock

§7.6.2.4.2 introduces a phase in which nothing is specified, so nothing is what
it produces — no group, no event, no action. It advances the offset the next
phase starts from, and the clock is already running.

A range duration (§7.6.2.4) constrains accepted traces rather than fixing a
time; choosing from it needs a solver (ADR-0004), so it is reported. §7.6.2.5's
emit, wait @event, on and until are reported too: those events are
abstract control objects with no runtime carrier in v0.0.1.

A pre-existing lexer defect this sprint surfaced

§7.2.2.6.7 spells the range constructor '[' expression '..' expression ']',
and §7.2.1.5.2's float-literal ::= digit* '.' digit+ makes the leading digits
optional — so [2..4] is a race between the .. operator and the float .4,
and the float was winning. The lexer emitted no .. token at all, and the
parser looked for ..., a spelling the standard does not have.

Because the two agreed with each other, every range in the test suite was
written the wrong way and nothing noticed. duration: [10s..30s] is what made
it visible. Fixed in both, with the affected test expectations corrected to the
standard's spelling and the race pinned in dsl_lexer_test.cpp.

Tests and docs

• dsl_lowering_test.cpp 37 → 45: absolute start times from durations,
  completion chaining without them, a nested parallel inside a serial, an
  AND-shaped join, wait elapsed, one_of default and named, an alternative
  that is not there, a range duration, and a non-default overlap.
• scena_run_test.cpp +2: --select choosing an alternative end to end, and a
  serial scenario whose phases take over at the second it named.
• dsl_lexer_test.cpp +2: the ..-versus-.4 race, in both plain and physical
  literals.
• ADR-0031; coverage-matrix rows for §7.6.2.1–§7.6.2.5 and §7.2.2.6.7; the
  scena-run user guide and the DSL frontend README.

1378 gtest (+12), 194 pytest.

Closes #45