Specialize inverse_cdf for LogNormal distribution

[Twister915]

Description

Specialize inverse_cdf for LogNormal distribution, because an equation exists and it is faster & more accurate than the default binary search implementation.

Changes

• Adds a definition for inverse_cdf in ContinuousCDF impl for LogNormal distribution, according to the definition I read here and some pen & paper verification of this definition.
• Generalizes the unit tests in LogNormal such that the CDF test cases are re-used (where applicable) to verify the inverse CDF definition

Review Notes

• I'm no mathematician so someone definitely needs to verify my implementation. I am not sure if the wolfram-alpha I linked above is the most general definition of the inverse CDF.
  • Wikipedia (seemingly) has a different definition which includes erf_inverse, and I don't know how erf_inverse relates to erfc_inverse.
  • I inverted the CDF definition myself (using my basic algebra skills) on pen & paper and it matched the wolfram-alpha math, so that's what I implemented.
• The unit testing section will perform the exact same test cases for the CDF, despite all the code changes. There is one exception: a change to the accuracy argument on the case where input=0.5 and output=0.0000000015011556178148777579869633555518882664666520593658.

[henryjac]

erfc is the complementary error function, so erfc(x) = 1 - erf(x). Wikipedia and wolfram alpha give the same math as your code, so that checks out. Will check the tests soon.

I like the way the tests are done so we check the cdf against the inverse_cdf, could highlight problems such as 200 more clearly.

[YeungOnion]

merge edits

• added } since added functions within impl ContinuousCDF for LogNormal were adjacent
• math in docs should be text instead of ignore with b8a0ec2

I too like the symmetric tests.

@henryjac iirc you wanted to rebase and merge so merge commits would represent PRs in history. Would you prefer that @Twister915 redo the changes atop master so we can rebase instead and keep that up?

[YeungOnion]

Opted to do a local merge and push to master, will have to close this one manually.

If anyone knows a way to manually modify this PR to "close as merged" by a specific commit, the SHA is ae5431f

Thanks for implementing this along with a thorough write-up!