Remove usage of fieldnames function at runtime where possible

[non-Jedi]

Introspection using fieldnames at runtime can have a large negative
effect on code speed.

For time-to-first-plot, which is all I'm interested in at the moment,
these changes give a 5-10% improvement on simple cases I tested.

Before:

const x = collect(1:2000)
const y = rand(2000)

@time p = plot(x=x, y=y, Geom.line);
 0.637450 seconds (947.25 k allocations: 48.127 MiB, 5.71% gc time)

@time draw(SVG("test.svg", 4cm, 4cm), p)
 60.589051 seconds (82.63 M allocations: 4.113 GiB, 5.98% gc time)

After:

const x = collect(1:2000)
const y = rand(2000)

@time p = plot(x=x, y=y, Geom.line);
 0.594075 seconds (941.17 k allocations: 47.927 MiB, 4.30% gc time)

@time draw(SVG("test.svg", 4cm, 4cm), p)
 54.662001 seconds (81.76 M allocations: 4.070 GiB, 5.74% gc time)

I didn't always get the full 10% increase shown here, but it was
consistently better across a few runs.

[codecov-io]

Codecov Report

Merging #1276 into master will decrease coverage by 0.25%.
The diff coverage is 60%.

@@            Coverage Diff             @@
##           master    #1276      +/-   ##
==========================================
- Coverage   90.33%   90.08%   -0.26%     
==========================================
  Files          36       36              
  Lines        3943     3953      +10     
==========================================
- Hits         3562     3561       -1     
- Misses        381      392      +11

Impacted Files	Coverage Δ
src/mapping.jl	84.46% <ø> (-0.98%)	⬇️
src/data.jl	0% <0%> (ø)	⬆️
src/scale.jl	97.41% <100%> (-1.28%)	⬇️
src/statistics.jl	96.84% <100%> (ø)	⬆️
src/aesthetics.jl	80.17% <60%> (-2.13%)	⬇️
src/theme.jl	71.18% <0%> (-1.23%)	⬇️
src/misc.jl	65.77% <0%> (-0.45%)	⬇️
... and 1 more

---

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 9df68de...ec2e1a3. Read the comment docs.

[bjarthur]

lgtm. passes regression tests i presume?

[non-Jedi]

I knew I was forgetting something. Hold on.

[non-Jedi]

Regression tests are okay. I actually got a positive, and then I
realized that the output of date_bar.jl depends on the day you run
it, and I was comparing against running the regression suite on master
yesterday. Will open a PR to make that one deterministic.