FEAT: Add `LQA` solution method

[QBatista]

Supercedes #31

Some key points regarding the implementation:

1. Use finite differences instead of forward mode AD to allow for f and g functions with array-like arguments restricted to floats.
2. Back out implied basis coefficients from the linear quadratic approximation. Although this method doesn't require a basis, this makes the output consistent across solution methods and makes it easier to chain solution methods.
3. The point around which the approximation is carried out is provided by the user. While the steady-state is a natural point for constructing the approximation, the user may be interested in doing the approximation at some other point. Besides, without imposing restrictions on f and g, it is not always clear what the value of the steady-state shock(s) should be.

[codecov[bot]]

Codecov Report

Merging #38 (6760602) into master (815177f) will increase coverage by 1.12%. The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master      #38      +/-   ##
==========================================
+ Coverage   89.25%   90.37%   +1.12%     
==========================================
  Files           2        2              
  Lines         214      239      +25     
==========================================
+ Hits          191      216      +25     
  Misses         23       23              

Impacted Files	Coverage Δ
src/cdp.jl	88.99% <100.00%> (+1.49%)	:arrow_up:

---

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 815177f...6760602. Read the comment docs.

[oyamad]

@QBatista Thanks!

• Run all the cells in the notebook.
• Please avoid using unicode character.

[oyamad]

@QBatista Thanks. Will you run through the notebook one more time (to get rid of Info: Precompiling ContinuousDPs ... from the 1st cell)?

[oyamad]

Please run all the cells...

[oyamad]

@QBatista Great work, thanks!