chore(exporters): delete dead otlphttp + stdoutexporter wrappers

[trilamsr]

Summary

The OCB build assembled from builder-config.yaml has shipped upstream go.opentelemetry.io/collector/exporter/otlphttpexporter + debugexporter since v0.2.0 (lines 59-60). The two in-tree exporter wrapper packages were left behind after the RFC-0013 distribution-first pivot and have had zero Go importers outside their own _test.go files since the pivot landed. This PR finishes that cleanup.

• Delete components/exporters/otlphttp/ — 13 files / 2,904 LOC. Superseded by upstream otlphttpexporter v0.130.0.
• Delete components/exporters/stdoutexporter/ — 8 files / 1,077 LOC. Superseded by upstream debugexporter v0.130.0 (component-id debug).
• Net: -21 source files / ~3,981 LOC; -4 collector deps (exporter, exporter/exportertest, indirect consumertest + xexporter); -42 lines off go.mod + go.sum.

Closes #333, #334.

Root cause

Not a bug — strategic dead code. Per the v1.0-rc1 simplification audit (docs/v1-rc1-simplification-audit.md §7 rows 1 + 2, risk score 1), both wrappers had:

• Zero external Go importers (only their own _test.go files).
• No reference from the OCB-generated _build/components.go.
• A live upstream supersession already wired through builder-config.yaml.

The followups shard docs/followups/otlphttp.md had every code-path item tagged [STRIKE] against the wrapper — that shard is now deleted along with the code it tracked.

Operator-impacting changes

- chore(exporters): delete in-tree `otlphttp` + `stdoutexporter` wrapper packages now that the OCB build ships upstream `otlphttpexporter` + `debugexporter` directly.
  - Configs naming `otlphttp` as an exporter type continue to work — the upstream component-id is identical.
  - Configs naming `stdoutexporter` should switch to `debug` (the upstream `debugexporter` component-id). The Helm chart default already uses `debug`.
  - PromQL queries on the wrapper's per-component `otelcol_exporter_otlphttp_calls_total` / `otelcol_exporter_stdoutexporter_*` instruments will go silent. Upstream emits the standard `otelcol_exporter_sent_*` / `otelcol_exporter_send_failed_*` shape via `exporterhelper` instead. See [`docs/migration/v0.1-to-v0.2.md`](docs/migration/v0.1-to-v0.2.md) "In-tree receiver / exporter namespace alignment" for the substitution annotations.

What else moved

Doc + workflow + config cascade so nothing dangles at a deleted package:

• module/pkg/selftel/selftel.go — package doc updated; the two wrappers no longer exist as "copies it replaced."
• .github/workflows/install-bench.yml — dropped components/exporters/otlphttp/** from pull_request + push paths triggers; header comment now says "upstream otlphttpexporter."
• bench/install/README.md — narrative updated to "upstream otlphttpexporter"; the failed-test-reference bullet now points at upstream tests.
• docs/FAILURE-MODES.md — two rows (network-error mid-send, bad-scheme endpoint) re-pointed from the in-tree wrapper tests to upstream otlphttpexporter / confighttp.
• docs/README.md — dropped the wrapper README index row; integration-recipe rows now say "upstream otlphttp exporter" instead of "in-tree."
• docs/STRATEGY.md — "queued for deletion" list no longer carries stdoutexporter; explicit note that both wrappers were deleted post-v0.2.0.
• docs/MILESTONES.md — M1 row's "the bundled stdoutexporter canonical example remains" annotation now reflects post-v0.2.0 retirement.
• docs/migration/v0.1-to-v0.2.md — per-component substitution table footnotes the wrapper retirement + instrument-name implication (the PromQL diff recipe in the file still shows the v0.1.x→v0.2.0 transition, which is correct history).
• docs/research/baselines.md — stdoutexporter coverage row dropped.
• docs/v1-rc1-simplification-audit.md — orphan-wrappers table now shows only pyspy (the remaining row); ranked-deletion candidates 1 + 2 marked Done with PR links.
• docs/v1-rc1-test-audit.md — coverage + integration-test rows dropped.
• CHANGELOG.md — [Unreleased] Removed entry added; two stale "source survives in-tree" / "eight surviving in-tree components" annotations corrected.
• docs/followups/otlphttp.md — deleted (every code-path item was [STRIKE] against the now-deleted wrapper).

Left untouched as historical decision-trail (per feedback_no_bloat's "archived files are suspect" being asymmetric — these are research-snapshot history, not active dead code):

• docs/research/m5-m6-research.md §F10 — original "ship components/exporters/otlphttp/ as in-tree adapter" decision text. RFC-0013 superseded this; the file as a whole carries the supersession context.
• docs/rfcs/archived/0004-clockreceiver-stdoutexporter.md — already archived.

Chart values + recipes use otlphttp as the upstream component-id (the chart key for the OCB-bundled exporter), not the wrapper import path — no change needed there.

Test plan

• git rm -r both directories; verify components/exporters/ removed (no empty parent).
• GOWORK=off go test ./... green at root.
• GOWORK=off go test ./... green at module/.
• GOWORK=off go vet ./... clean.
• make tidy drops exporter, exporter/exportertest, consumertest, xexporter; make tidy-check clean afterwards.
• ./scripts/register-lint.sh clean (no func Register* outside components/).
• make coverage-check clean (pyspy 82.3% ≥ 60% gate).
• make doc-check + chart-appversion-check.sh clean.
• Pre-commit gates (lint, vet, mod-verify, attribute-namespace-check, no-autoupdate-check) all green.
• CI install-bench workflow re-runs against the trimmed paths trigger (will fire on first PR that touches the listed paths post-merge).
• CI full make ci + install-bench pass on this PR.

[trilamsr]

Addressed reviewer nit: dropped two stale 'in-tree wrapper retired post-v0.2.0' footnotes in docs/FAILURE-MODES.md (lines 96, 99). Test column now cites upstream paths directly. Historical context lives in CHANGELOG. (Reviewer cited three; only two occurrences existed in-file.)