A pair, and a point, as values of their own, distinct from a vector #2149

Modenov's Problems in Geometry (#2131) works with points, and a point is not the same thing as a vector or a complex number:

• Chapter I treats a directed segment as a vector.
• Chapter III treats a point as its complex affix.
• Chapter II writes coordinates (x, y), and affine and oblique frames.

AngouriMath has no point or pair today:

• [1, 2] is a column vector.
• [1, 2] = [1, 3] gives [True, False] (A / A for a matrix is computed elementwise, and so is the undefined product of two columns #2141).
• [1, 2] in RR^2 checks each coordinate separately.
• A set over pairs, { [x, y] : … }, parses but never decides membership.
• (a, b) does not parse at all, so it is free to become a pair.

The proofs-book reading (#1409, chapter 3) found the ordered pair and the Cartesian product missing too.

Already decided for v3 (Happypig375):

• On Pending breakages: changes agreed or arguable that cannot ship before a major version #1019: [a; b] intervals become [a, b], "interpreted as intervals where sets are expected and vectors otherwise", and ; becomes the line separator.
• On Let statement #352: points are a type, as in let A, B, C : point.

Wanted, within those decisions:

• a pair (tuple) node, written (a, b), read as an open interval where a set is expected and as a pair otherwise, by the same rule, with RR × RR as its set;
• point as Let statement #352's type, and what the book's three readings of a point (a vector, a complex affix, a coordinate pair) map to.

Lines, circles and loci as sets of such points then follow (intersection already exists for sets). That is why this is 3.0 and the geometry built on it is 3.2.

Related: #330 (sets by condition over more than one variable), #1237 (typed variables).