feat: add benchmark filtering (--max-size, --category, --skip-longctx)

Both run-baseline.sh and run-suite.sh now support:
- --max-size GB: skip models larger than N GB (prevents OOM)
- --category LIST: filter by catalog category (smoke,dense,moe)
- --skip-longctx: skip 32K context tests (saves time + memory)
- --reps N: configure repetition count
- --help: shows usage with examples

Safe pre-optimization run: benchmark baseline --max-size 20 --skip-longctx
Full post-optimization: benchmark baseline (no filters, all models + longctx)

Also: 4 new BATS tests for flag parsing (98 total, all passing)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

tests/benchmark_flags.bats

@@ -0,0 +1,37 @@
+#!/usr/bin/env bats
+# Tests for benchmark script filtering flags
+
+load test_helper.sh
+
+@test "baseline --help shows usage and exits 0" {
+    run bash "$PROJECT_ROOT/scripts/benchmark/run-baseline.sh" --help
+    assert_success
+    assert_output --partial "Usage"
+    assert_output --partial "--max-size"
+    assert_output --partial "--category"
+    assert_output --partial "--skip-longctx"
+}
+
+@test "run-suite --help shows usage and exits 0" {
+    run bash "$PROJECT_ROOT/scripts/benchmark/run-suite.sh" --help
+    assert_success
+    assert_output --partial "Usage"
+    assert_output --partial "--max-size"
+    assert_output --partial "--category"
+    assert_output --partial "--skip-longctx"
+    assert_output --partial "--tag"
+}
+
+@test "benchmark dispatcher shows help with no args" {
+    run bash "$PROJECT_ROOT/bin/benchmark"
+    assert_failure
+    assert_output --partial "Commands"
+    assert_output --partial "--max-size"
+    assert_output --partial "--skip-longctx"
+}
+
+@test "benchmark dispatcher passes --help through to baseline" {
+    run bash "$PROJECT_ROOT/bin/benchmark" baseline --help
+    assert_success
+    assert_output --partial "Usage"
+}