fix(cddl): preserve float literals so (0.0..1.0) is distinct from (0..1)

[titusfortner]

What does this PR do?

• Numeric literal nodes in the parsed AST now record whether the value was written as a float, via a new IsFloat: true marker on float literals.
• A floating-point range like (0.0..1.0) is now distinguishable from an integer range (0..1). Previously both produced byte-for-byte identical nodes (Min.Value: 0, Max.Value: 1), because whole-valued floats such as 1.0 collapsed to plain integers — the lexer's FLOAT vs NUMBER distinction was lost at AST construction.
• Consumers walking the AST (validators, code generators, downstream tooling) can now tell 1.0 from 1 without re-lexing.

Implementation Notes

• The marker is only present on float literals (IsFloat?: boolean, absent for integers), so the change is purely additive — existing integer literal nodes are unchanged and no consumer relying on the current shape breaks.
• Range Min/Max inherit the marker for free, since both are built from these literal nodes.
• Scope kept to the parser (the shared root cause). The bundled CDDL→language generators (cddl2java, cddl2py, cddl2swift, cddl2kotlin) emit an integer type for float ranges and can now consult IsFloat to fix that — tracked separately. cddl2ts needs no change (TypeScript has a single number type).
• Snapshot updates are limited to genuine float literals already present in fixtures (e.g. 0.1, 1.5) gaining the new marker.

[christian-bromann]

LGTM 👍