CBL-6626: SQL++ query not return expected results

[callumbirks]

The issue which caused the customer's query to not return the correct results is that litecore thinks the null value is not null.
The underlying cause of that is an issue with indexes which causes some values in certain indexes to not be correctly typed as null when passing into the query engine.
So this does not fix the underlying problem, but it is a good workaround that @snej agreed with.
Basically the problematic values do not get given the Null subtype when they should, so our usual check for the null subtype fails. Now, we also check if the size is 0 if the subtype has not been set.

[cbl-bot]

Code Coverage Results:

Type	Percentage
branches	67.94
functions	79.23
instantiations	35.25
lines	79.28
regions	75.17

[jianminzhao]

@snej I see Blob&&subtype==0 is being used in fl_result. I would defer to you whether to accept it as a patch to isNull() function.

[jianminzhao]

Please also fix fl_result where we only check blob's subtype but not the size. Also for isNull(), size of the blob (being 0) is sufficient. Subtype is secondary. (non-zero size with subtype fleecenull reveals a bug)

[callumbirks]

@jianminzhao I've added the fix to fl_result.

I'm unsure about removing the check for subtype == 0, this is what @snej told me was appropriate previously.
There are more subtypes which are used with blob, i.e. kPlainBlobSubtype, which is encoded as a slice, so may be allowed to be an empty slice - this I am unsure of.

[jianminzhao]

Is it equivalent to:
case 0:
case kFleeceNullSubtype:
if (size > 0) break; // untagged blob ...
else {
// Fleece/JSON null
return;
}