Define & implement improvements to analysis when no data present

When there is no data present for lane / device, margin_summary fails immediately.

We should 1. handle these errors so we still analyze other lanes and 2. gather to sufficient context to understand why there was no data.

Example:
https://mfg.oxeng.dev/activities/01KCPAMD9JHHCKKN07GWTNSJTV#event-342

/opt/oxide/lmar/lib/margin_summary.py:169: UserWarning: loadtxt: input contained no data: "/tmp/tmp.wYaO6M/unpack/margin-1986-12-28T00-00-23/margin-results-b21-d0-f0-l0"
  results = np.loadtxt(file, skiprows=4, delimiter="\t")
Traceback (most recent call last):
  File "/opt/oxide/lmar/lib/margin_summary.py", line 689, in <module>
    main(sys.argv[1:])
  File "/opt/oxide/lmar/lib/margin_summary.py", line 685, in main
    direct_mode(ns.inputs, ns.outdir, ns.out, ns.pass_count_required, ns.pass_err_cnt, symmetric=ns.sym_eye_calc)
  File "/opt/oxide/lmar/lib/margin_summary.py", line 638, in direct_mode
    text = summarize(files, pass_count_required, pass_err_cnt, symmetric=symmetric)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/oxide/lmar/lib/margin_summary.py", line 216, in summarize
    results = load_results(file, pass_err_cnt=pass_err_cnt)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/oxide/lmar/lib/margin_summary.py", line 171, in load_results
    is_time = results[:, time] != 0
              ~~~~~~~^^^^^^^^^
IndexError: too many indices for array: array is 1-dimensional, but 2 were indexed
ERROR: lmar-margin-summary failed.

Start with load_results and summarize in margin_summary.py, using the traceback and linked activity as the reproduction context. Follow how an empty lane or device result reaches the array indexing, then verify that analysis continues for other lanes and that the no-data case reports enough context to explain it.