Avoid missing cube files on reading

[osopardo1]

Due to the bug stated in issue #47 and according to reasoning in #48, we should work on avoiding losing cube records at reading time.

The greedy estimation in the first step of the writing protocol causes having missing parents in the index structure. This means that all data is written down, but the read operation is not supposed to contemplate a behavior in which cube "AAAA" is present but parent "AAA" is not. If a father is not present in the index structure because of bad index quality, the query should be anyways retrieving all the necessary elements.

For that, this Draft PR is for:

• Use Patricia Trie structure instead of Map for resolving the CubeWeights. A Patricia Trie would find for any matches in a CubeId or its descendants. => No need, a SortedMap is enoguh
• Possibly change the recursive implementation to an iterative one. This would help also with issue Transform recursive method findSampleFiles to iterative #56

[osopardo1]

Hey @Jiaweihu08 @cugni you could check this one for the pair programming session today 😈

[cugni]

Why do we need the previouslyMappedFiles? What will delta filter that we cannot? Is there a specific case? if not, it would be better just to have the QbeastFiles directly to the indexStatus, instead of intersecting the two lists. Also, OTreeIndex, we are intersecting the file list twice, which is unnecessary.

[cugni]

It seems good!

[cugni]

We still need to ensure that the QueryExecutor skips the files with maxWeight < weightRange.from

[eavilaes]

When creating a new test in QueryExecutorTest.scala, I found out that most of these tests are using a cubeSize = 10.
This leads to the cubeSize being ignored and the config.DEFAULT_CUBE_SIZE is used.
Is this the desired behavior for these tests? There's already a specific test that checks the behavior using a smallCubeSize.