Support scientific notation for float literals (1e-12, 2.5e3)

[timfennis]

Summary

The lexer does not support scientific/exponential notation for float literals. 1e-12 currently causes a parse error.

Example

let eps = 1e-12;   // parse error: Expected ';' but got 'e'
let big = 2.5e3;   // also fails

Motivation

Came up when writing tests for trig functions — a natural epsilon value like 1e-12 had to be written as 0.000000000001, which is error-prone and hard to read.

Expected behavior

Scientific notation should be parsed as float literals:

• 1e12 → 1000000000000.0
• 1e-12 → 0.000000000001
• 2.5e3 → 2500.0
• 1.5E-6 → 0.0000015 (case-insensitive E)

Implementation hint

Update the float literal rule in ndc_lexer to accept an optional exponent part ([eE][+-]?[0-9]+).