Precompute row and column counts

[dmbates]

• At the time that a .bed file is memory-mapped, compute the column and row counts
• Provide views of the column and row counts as Vector{SVector{4,Int}}
• Modify summary statistics (mean, var, maf, minorallele, missingrate) to use the cached column or row counts.
• I attempted to create a new version of the grm calculation but it didn't go well. The original from @Hua-Zhou performs better. Mine is called newgrm and not exported. For some reason it allocates about 10 times as much memory as the earlier version.
• It is not clear that the work on precomputing the row and column counts is worthwhile. For one thing, they only work properly with files opened read-only. If an element of the data array is modified the cached values will no longer be valid.

[coveralls]

Pull Request Test Coverage Report for Build 27

• 58 of 115 (50.43%) changed or added relevant lines in 3 files are covered.
• 1 unchanged line in 1 file lost coverage.
• Overall coverage increased (+1.9%) to 27.857%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/bedfile.jl	8	19	42.11%
src/summarystats.jl	50	68	73.53%
src/grm.jl	0	28	0.0%

Files with Coverage Reduction	New Missed Lines	%
src/bedfile.jl	1	14.75%

Totals
Change from base Build 18:	1.9%
Covered Lines:	78
Relevant Lines:	280

---

💛 - Coveralls

[codecov]

Codecov Report

Merging #2 into master will increase coverage by 1.94%.
The diff coverage is 50.43%.

@@            Coverage Diff             @@
##           master       #2      +/-   ##
==========================================
+ Coverage   25.91%   27.85%   +1.94%     
==========================================
  Files           2        4       +2     
  Lines         247      280      +33     
==========================================
+ Hits           64       78      +14     
- Misses        183      202      +19

Impacted Files	Coverage Δ
src/BEDFiles.jl	100% <ø> (ø)	⬆️
src/grm.jl	0% <0%> (ø)
src/bedfile.jl	14.75% <42.1%> (-10.86%)	⬇️
src/summarystats.jl	73.52% <73.52%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ddd7939...e14e5f7. Read the comment docs.