Improve performance of .NET Core Analyze

[johnduhart]

Calling file.Sync() for each log message has an extremely negative performance impact. Additionally, logging each Package/Project reference in a mid-sized solution will result in a 2GB log file.

There's likely some additional perf gains to be had in this analyze implementation, but this is the low-hanging fruit.

[CLAassistant]

All committers have signed the CLA.

[elldritch]

PR LGTM. Thanks for the contribution! We'll merge once we get the tests passing (this looks like CI misconfiguration on our end, we'll take care of it).

@zlav can you take a look at the CI failures here? They're related to the coverage statistics we added.

It looks like we might want to do something similar to cloudflare/cfssl#203.

[elldritch]

I'm a bit concerned that panics and log.Fatals might not flush the log if we get rid of per-entry sync'ing.

To improve the performance of this, we should probably have a defer func() { logFile.Sync() } in main. We'll need to migrate away from log.Fatal usage before that, since log.Fatal calls os.Exit which bypasses deferred functions.

[johnduhart]

@liftM I put the sync back in, it provided no perf benefit when the log message in question was removed.

[elldritch]

Merging despite end-to-end test failure so I can cut a release.

@lackstein: this end-to-end test failure is because push-only keys still do not work correctly, so we're unable to do tests with API keys for public forks. What's the status on that?

[coveralls]

Coverage remained the same at 35.631% when pulling fc4d0c7 on mdsol:fix/net-performance into e9ac753 on fossas:master.