perf: Add bounds for haplotype matrix

We know that the upper bound for the value of any cell in the occurence
matrix has to be equal to or less than the number of reads. Leverage that
knowledge to shrink the size of arrays that are needed.

src/haplotypecalling.jl

@@ -253,7 +253,16 @@ dimensional matrix.
 function occurrence_matrix(
     haplotype::AbstractArray{Variation{S,T}}, reads::AbstractArray{Haplotype{S,T}}
 ) where {S<:BioSequence,T<:BioSymbol}
-    hapcounts = SparseArray{UInt}(undef, Tuple(repeat([2], length(haplotype))))
+    Q = UInt
+    for int_type in [UInt8, UInt16, UInt32, UInt64, UInt128]
+        if length(reads) < typemax(int_type)
+            Q = int_type
+            break
+        end #if
+        error("Too many reads to represent in memory")
+    end #for
+
+    hapcounts = SparseArray{Q}(undef, Tuple(repeat([2], length(haplotype))))
 
     for read in reads
         coordinates = zeros(Int, size(haplotype))