Provide way to stop normalization if the expression is obviously problematic (such as deletions in large gap/unknown regions)

When trying to normalize the variant `NC_000015.9:g.7211_7214del` the routine will go into a seemingly endless routine to try to figure out the `normalized` result for the `Allele.state`.

Without a full analysis their is evidence that this is likely caused by the fact that the first 17 million bases in chromosome 15 are all `N`s. So as it rolls right/left to get to a unique sequence region it will go on for an impractical amount of time.

I suggest we put a limit in terms of how large the sequence can grow up to when normalizing the Allele. But we should discuss how to best handle this.

@toneillbroad just suggested that maybe we simply disallow any normalization that includes ambiguity coded bases not `A, C, T or G`. I sort of like that as a general rule of thumb, since it is very difficult to address the true `normality` of a sequence that includes any of the ambiguity codes. We can make this a vrs-python rule so that our normalizer doesn't go off and never return in these portions of the reference sequences