Go: mixing int and float operands takes the first operand's type, so ready ? 1 : 2.5 or x := 1 + 0.5 fails to compile #152

What's wrong

Static type inference in StandardLanguageGenerator (Coder/Languages/StandardLanguageGenerator.cs:160-161) types:

• a BinaryExpression as StaticType(Left) ?? StaticType(Right);
• a conditional as the type of its first branch.

There is no numeric promotion. GoGenerator then uses that type for the closure it emits for a conditional (GoGenerator.cs:1658) and for the := local registry.

Reproduction (go build, HEAD ad68d48)

A function returning double contains return abs(ready ? 1 : 2.5). It generates:

abs(func() int { if ready { return 1 }; return 2.5 }())

go build reports:

• cannot use 2.5 (untyped float constant) as int value in return statement (truncated)
• cannot use func() int {…}() (value of type int) as float64 value

x := 1 + 0.5 is likewise registered as int, which breaks later typed uses of x.

This came in with e2c5afa ("Go type inference"). It is distinct from #134, which is about the nested-conditional closure shape.

Suggested fix / acceptance criteria

• When combining operand or branch types, apply numeric promotion: float64 over float32 over int64 over int. For example, a ResultType(a, b) helper.
• For the conditional closure, prefer a known target type (the enclosing return type or parameter type) over the inferred branch type.
• Add Go compile tests for a mixed int/float conditional used as an argument and as a return value, and for x := 1 + 0.5 followed by a float64 use.