Sort metrics alphabetically in EXPLAIN ANALYZE output

[progval]

Which issue does this PR close?

Closes #12567.

Rationale for this change

From ParquetExec metrics, with predicate pushdown enabled:

metrics=[output_rows=0, elapsed_compute=96ns, row_groups_matched_bloom_filter=0, row_groups_matched_statistics=21050, file_scan_errors=0, pushdown_rows_matched=0, row_groups_pruned_statistics=173576, row_groups_pruned_bloom_filter=21050, file_open_errors=0, num_predicate_creation_errors=0, bytes_scanned=25023432248, pushdown_rows_pruned=0, page_index_rows_pruned=0, predicate_evaluation_errors=0, page_index_rows_matched=0, time_elapsed_scanning_until_data=16.622753ms, time_elapsed_processing=72.280463608s, pushdown_eval_time=382ns, page_index_eval_time=3.177676ms, time_elapsed_scanning_total=16.661811ms, time_elapsed_opening=102.989073638s]

For example, pushdown_rows_matched and pushdown_rows_pruned (highlighted in the snippet) are far away from each other, even though they refer to roughly the same thing.

The unstable sort also makes it hard to compare multiple EXPLAIN ANALYZE results.

after this change, metrics for the same query look like this:

metrics=[output_rows=0, elapsed_compute=96ns, bytes_scanned=25023432248, file_open_errors=0, file_scan_errors=0, num_predicate_creation_errors=0, page_index_rows_matched=0, page_index_rows_pruned=0, predicate_evaluation_errors=0, pushdown_rows_matched=0, pushdown_rows_pruned=0, row_groups_matched_bloom_filter=0, row_groups_matched_statistics=21050, row_groups_pruned_bloom_filter=21050, row_groups_pruned_statistics=173576, page_index_eval_time=2.882359ms, pushdown_eval_time=382ns, time_elapsed_opening=104.629010525s, time_elapsed_processing=73.86660138s, time_elapsed_scanning_total=97.929057ms, time_elapsed_scanning_until_data=97.893962ms]

What changes are included in this PR?

Refinement of the partial order used in MetricsSet::sorted_for_display

Are these changes tested?

Yes

Are there any user-facing changes?

More readable output. There doesn't seem to be any snippet in the documentation that needs to be updated to match the new behavior.

[alamb]

Thank you @progval -- this is a very nice improvement in my mind

[Weijun-H]

LGTM! Thanks @progval