Also define merc (resp. tmerc) as the conformal projection in which the
equator (resp. a chosen meridan} projects to a straight line at constant
scale.

This addresses item 1 in issue #2387

Things to note:

* I made "editorial" changes to the text.  The virtues and vices of
  Mercator are a hot topic.  So check these out.  (I judged that the
  text I replaced to be pretty misleading.)

* I include the radius of the sphere/ellipsoid in the formulas (and I
  did this also for my mods for tmerc documentation).  Surely this is
  better than leaving the reader to figure out how this is introduced.

* I include the "old-style" (ca 18th century) formulas and the newer
  ones in terms of hyperbolic functions.  The former may be the familiar
  ones, but the latter are better for computation (more succinct, more
  accurate, faster, preserve parity).

* For the inverse ellipsoidal transformation, I just say that the
  formula for psi is inverted iteratively.  This is probably sufficient,
  but it could be expanded later.

Instead of 'elliptical'.
Was suggested by @cffk in https://github.com/OSGeo/PROJ/pull/2030#issuecomment-596132341

An elliptical projection is a projection whose global shape fits inside an
ellipsoid, like Mollweide. At least according to
https://www.merriam-webster.com/dictionary/elliptical%20projection
and
https://en.wikipedia.org/wiki/Mollweide_projection

This had unexpected side-effects when building the HTML docs on Travis
compared to locally. The correct solution would be to set different
scaling depending on the output.

This reverts commit 325477ed0fac2c9233c2f6a2b7bb4125e04df24c.