argv dissolution: replace hand-typed id/hostname argv with typed builders (host_reach_identity_probe)

[gunbai-bot]

Summary

Replaces two hand-typed argv arrays in gunbc.host.host_reach_identity_probe with typed builders from extdeps.tools.*:

Before	After
["id", "-un"]	id_user_name_argv() (new builder in extdeps.tools.id)
["hostname"]	hostname_short_read_argv() (existing builder in extdeps.tools.hostname)

Changes

1. dag/extdeps/tools/id.dag: Added id_user_name_argv() producing ["id", "-u", "-n"] — the decomposed form equivalent to ["id", "-un"] for printing the effective username (name, not number).

2. dag/gunbc/host/host_reach_identity_probe.dag: Updated imports and both call sites to use the typed builders.

Why this matters (argv dissolution)

Per transport_argv_anemia_dissolution and DP-M2 roadmap item "Replace hand id/hostname argv with builders":

• Flags become modeled axes — success semantics explicit in the builder, making omit-by-typo unwritable
• argv is derived from the model, not the reverse
• Typed argv executes directly via typed_argv_probe_over_ssh — no re-joining into shell text that discards the checking the typing bought
• Single source of truth for id(1) / hostname(1) flag conventions

Intentional behavioral improvement

probe_ready_target_once now runs hostname -s instead of bare hostname. This is intentional, not a refactor side-effect:

• The result is stored in ProbeReachable.observed_short_hostname — the field name asks for a short hostname
• On hosts where plain hostname prints the full FQDN, the probe now correctly gets the short name
• This matches the field's semantic contract and is a fix, not a regression

Verification

• cargo check ✅
• Unit tests pass (1098 passed, 6 pre-existing failures unrelated) ✅
• host_reach_identity_probe_witness_test passes ✅

[gunbai-bot]

Closed as superseded by #13626 (folded into #13641): both replace the hand-typed id/hostname argv in host_reach_identity_probe with the extdeps.tools builders. Branch kept. — sent from neat-wolf-604