Improve metrics (and some other things)

[dominiklohmann]

This PR is based on @Dakostu's excellent #3376 that laid the groundwork for metrics.

It comes with the following metrics-related changes:

• We now calculate the approximate number of bytes in execution nodes when they handle events. The calculated number is an underestimate and (1) never includes the byte size of the schema, (2) excludes some array data buffers depending on the Arrow version used.
• tenzir --dump-metrics '<pipeline>' prints metrics on stderr after the pipeline finished.
• A set of bug fixes regarding metrics collection that were still in Expose pipeline operator metrics in execution node and pipeline executor #3376.

I also snuck in the following change that is unrelated to metrics, but that I stumbled over while developing this PR:

• Sorting now works for ip and enum fields.

This is what --dump-metrics looks like:

❯ # Using the M57 trace from the Get Started guide:
❯ cat Zeek/*.log | tenzir --dump-metrics 'read zeek-tsv | to file /dev/null'
operator #1 (load)
  elapsed: 4.02s
  scheduled: 553.13ms (13.75%)
  running: 484.85ms (12.05%)
  outbound:
    bytes: 199,898,907 at a rate of 49677650.77/s
    batches: 12,201 (16383.81 bytes/batch)
operator #2 (read)
  elapsed: 6.06s
  scheduled: 1.84s (30.34%)
  running: 1.78s (29.39%)
  inbound:
    bytes: 199,898,907 at a rate of 32995890.01/s
    batches: 12,201 (16383.81 bytes/batch)
  outbound:
    events: 953,677 at a rate of 157416.68/s
    bytes: 1,997,471 at a rate of 329708.32/s (estimate)
    batches: 54 (17660.69 events/batch)
operator #3 (write)
  elapsed: 7.96s
  scheduled: 7.67s (96.45%)
  running: 7.66s (96.24%)
  inbound:
    events: 953,677 at a rate of 119873.34/s
    bytes: 1,997,471 at a rate of 251074.02/s (estimate)
    batches: 54 (17660.69 events/batch)
  outbound:
    bytes: 541,837,222 at a rate of 68106745.97/s
    batches: 570 (950591.62 bytes/batch)
operator #4 (save)
  elapsed: 7.96s
  scheduled: 6.31ms (0.08%)
  running: 3.94ms (0.05%)
  inbound:
    bytes: 541,837,222 at a rate of 68102061.06/s
    batches: 570 (950591.62 bytes/batch)

[Dakostu]

Thank you for the addition, and the sort fix seems to come in just at the right time 💪

I left some minor comments.