Harden nvbench-compare plotting and filter docs

Skip UNKNOWN rows when collecting summary plot entries so non-numeric
fractional differences cannot reach the plotting path. Add a regression test
that exercises compare_benches(..., plot=True) with an UNKNOWN row.

Document the supported pow2 axis-filter syntax and update the CLI help example
to use NAME[pow2]=EXP, matching the parser behavior for axes displayed as 2^N.

* Document when status ???? (UNKNOWN) is emitted
* Clarify --no-color behavior

* nvbench_compare.md: clarify --no-color behavior, fix example

* Document display options in nvbench_compare.md

* Small mention of plotting capabilities in nvbench_compare.md
* Call out that example requires shell with process substitution capabilities

python/scripts/nvbench_compare.py

@@ -2807,7 +2807,11 @@ def compare_benches(
                                 comparison=comparison,
                             )
                         )
-                    if plot:
+                    if (
+                        plot
+                        and comparison.frac_diff is not None
+                        and math.isfinite(comparison.frac_diff)
+                    ):
                         axis_label = format_axis_values(axis_values, axes, axis_filters)
                         if axis_label:
                             label = f"""{cmp_bench["name"]} | {axis_label}"""
@@ -3038,7 +3042,7 @@ def main() -> int:
         dest="filter_actions",
         action=OrderedBenchmarkFilterAction,
         help=(
-            "Filter on axis value, e.g. -a Elements{io}=2^20. Applies to the "
+            "Filter on axis value, e.g. -a Elements{io}[pow2]=20. Applies to the "
             "most recent --benchmark, or all benchmarks if specified before any "
             "--benchmark arguments."
         ),

python/test/test_nvbench_compare.py

@@ -2117,6 +2117,39 @@ def test_compare_benches_counts_missing_summaries_as_unknown(
     assert row[-1] == "\U0001f7e1 ????"
 
 
+def test_compare_benches_plot_skips_unknown_rows(monkeypatch, nvbench_compare):
+    plotted_entries = []
+
+    def fake_plot_comparison_entries(entries, *args, **kwargs):
+        plotted_entries.extend(entries)
+        return 0
+
+    monkeypatch.setattr(
+        nvbench_compare, "plot_comparison_entries", fake_plot_comparison_entries
+    )
+
+    run_data = make_comparison_run_data(nvbench_compare)
+    ref_state = make_state(nvbench_compare, "state")
+    del ref_state["summaries"]
+    cmp_state = make_state(nvbench_compare, "state", mean="1.0")
+
+    nvbench_compare.compare_benches(
+        run_data,
+        [make_benchmark([ref_state])],
+        [make_benchmark([cmp_state])],
+        threshold=1.0,
+        plot_along=None,
+        plot=True,
+        dark=False,
+        filter_plan=make_filter_plan(nvbench_compare),
+        no_color=True,
+    )
+
+    assert run_data.stats.config_count == 1
+    assert run_data.stats.unknown_count == 1
+    assert plotted_entries == []
+
+
 def test_compare_benches_defaults_to_interval_display(monkeypatch, nvbench_compare):
     run_data = make_comparison_run_data(nvbench_compare)
     tabulate_calls = capture_tabulate_calls(monkeypatch, nvbench_compare)