[lmi] Testing various outputs

[Greg Chicares]

Soon we'll need to refactor 'emit_ledger.?pp', to accommodate a new
group premium report. I wrote these commands for testing:

/opt/lmi/bin[0]$cp -a /lmi/src/lmi/sample.cns /lmi/src/lmi/sample.ill .

/opt/lmi/bin[0]$./lmi_cli_shared --file=sample.ill --accept --ash_nazg 
--data_path=/opt/lmi/data \
 
--emit=emit_pdf_file,emit_test_data,emit_text_stream,emit_custom_0,emit_custom_1
 2>&1 |less -S

/opt/lmi/bin[0]$./lmi_cli_shared --file=sample.cns --accept --ash_nazg 
--data_path=/opt/lmi/data \
 
--emit=emit_pdf_file,emit_test_data,emit_spreadsheet,emit_group_roster,emit_text_stream,emit_custom_0,emit_custom_1
 \
 2>&1 |less -S

but found that the "custom" emissions overwrote the input files.
Fixed 20150803T1704Z, revision 6229.

Here's something else that could be regarded as an anomaly. The last two
commands above (for 'sample.ill' and 'sample.cns') both write
  'sample.test0' (emit_custom_0)
and
  'sample.test1' (emit_custom_1)
That means we can't run both and then test all the output--instead, we
must run and test each individually. This could be avoided easily: just
form output names by appending a new extension without dropping the old.
That affects regression testing, which we could work around by
renaming these output files in /opt/lmi/test/ :
  ini0*.test*   --> ini0*.ini.test*
  inix0*.test*   --> inix0*.inix.test*
Previously, 'sample.ill' and 'sample.cns' both produced 'sample.test1';
'sample.ill.test1' becomes distinguishable from 'sample.cns.test1' (and
likewise for '.test0') with such a change. But I don't think it best to
change that, notably because one broker has and uses the applicable part
of our regression-test suite, and we can't guess how much this change
might cost them. At any rate, '.test0' and '.test1' output by their
nature are intended to operate with invariant names controlled by class
configurable_settings:
    custom_input_0_filename --> custom_output_0_filename
    custom_input_1_filename --> custom_output_1_filename
where those four names are fixed constant strings, and the output files
are deliberately overwritten on each run...thus, what is deliberate may
be deemed not to be anomalous.