Updated geolocate speed

[Arkoniak]

Closes #52

With this PR, speed has improved significantly. @btime shows 27 ns which is 10^7 times faster than the current implementation.
Unfortunately, this comes at a cost of ~9Gb of geodata and almost a minute of data loading.

So, before merging, issues with extra memory occupation should be solved.

[codecov-commenter]

Codecov Report

Merging #53 (32ebeaa) into master (66cc4a7) will increase coverage by 0.58%.
The diff coverage is 94.00%.

@@            Coverage Diff             @@
##           master      #53      +/-   ##
==========================================
+ Coverage   87.83%   88.42%   +0.58%     
==========================================
  Files           2        2              
  Lines          74       95      +21     
==========================================
+ Hits           65       84      +19     
- Misses          9       11       +2     

Impacted Files	Coverage Δ
src/geoip-module.jl	89.28% <91.30%> (-0.19%)	⬇️
src/data.jl	88.05% <96.29%> (+0.78%)	⬆️

---

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 66cc4a7...32ebeaa. Read the comment docs.

[Arkoniak]

With 0a0375b update, I have the following benchmarks:

using BenchmarkTools
using GeoIP
using StableRNGs
import Sockets: IPv4, @ip_str

db = load(zipfile = "GeoLite2-City-CSV_20210427.zip");

rng = StableRNG(2021)
smp = rand(rng, db.index, 100)
ips = map(smp) do net
    IPv4(net.netaddr + 1)
end

ip = ips[1]    # ip"201.186.185.1"
ipnet = IPv4Net(ip, 32)

julia> @btime geolocate.($db, $ips);
  392.947 μs (4144 allocations: 284.28 KiB)

julia> @btime geolocate($db, $ip)
  2.947 μs (41 allocations: 2.84 KiB)
Dict{String, Any} with 21 entries:

It still can be Improved, since lookup takes only ~40 ns, so most of the 2.947μs time is spent on building a dictionary.