Remove assertions and skip on missing or broken rrname, bailiwick, rdata on nmsg files

[regalk13]

Replaces fatal assert() crashes on invalid records (missing or empty rrname/bailiwick/...) with per-record skip so conversion continues past bad data. And add count_skipped counter reported alongside count_messages.

[shw700]

I would suggest changing this to

if (((count_messages + count_skipped) % STATS_INTERVAL) == 0)

but otherwise, everything looks good

[cmikk]

Although I'm on board with replacing these assert() calls, this change risks silently discarding indicators of a potentially serious upstream issue.

Preferable alternatives would be:

1. exit with a failure code rather than abort() on bad data,
2. count bad data, and exit with a failure code if count_skipped > 0, or
3. option 1, but with logging.

I would be open to making "skip and count invalid messages" a configurable non-default behavior.

[kadealicious]

Although I'm on board with replacing these assert() calls, this change risks silently discarding indicators of a potentially serious upstream issue.

Preferable alternatives would be:

1. exit with a failure code rather than abort() on bad data,
2. count bad data, and exit with a failure code if count_skipped > 0, or
3. option 1, but with logging.

I would be open to making "skip and count invalid messages" a configurable non-default behavior.

I vote for option #3 with the option to skip+count invalid messages.

[cmikk]

This should be initialized -- C does not automatically initialize variables, so this could be true / nonzero leading to improperly skipped data.

[cmikk]

The default behavior should be to exit cleanly on invalid input, or break; here, so we don't continue processing when we've decided the input is not 100% valid data.

[cmikk]

From a style perspective, count_skipped > 0 is preferable to count_skipped >= 1 here.