Use of exceptions for flow control in BlockStore sublasses cause tests (and maybe other things) to take approx 200x as long when debugging

[drmason789]

B60D06DA-6A28-45F4-A074-94EC03A7C973.MP4

In the attached video, TestMetaDataIPI.TestFour and TestSix take 144ms and 110ms respectively when run.
When debugged, they take 25.3s and 28.9s respectively.

What's taking up all this time? When I do the crudest form of profiling, i.e. break and see what it's doing, I find it in implementations of BlockStore.CreateBlockIfNeeded.

This function is trying to get the block, creating a new one if an IndexOutOfBounds exception is thrown.

To mitigate this, introduce BlockStore.TryGetBlockAt(int offset, out ByteBuffer buffer) and make that avoid throwing IndexOutOfRangeExceptions, instead returning false when it can't find the block and true when it does.

This cuts the execution time of these tests down to about 40% of before.

Additionally, there is code in DirectoryNode.GetEntry that is catching a KeyNotFoundException and throwing a new, more informative exception when the key is missing. This exception handling can be avoided using _byname.TryGetValue(...).

This takes the test execution time down to about 20% of before.

[drmason789]

Why is help wanted?

[tonyqus]

Sorry, I think this is a mistake after long-time delaying. I forgot this issue is fixed by the PR.