deny unsafe math operations

[sam0x17]

Part of #300, we need to deny the ability to perform potentially-panicking and/or overflowing operations on number types in subtensor

AC:

• [ ] fix existing offending overflow operations
• [ ] fix existing offending float usages, if any
• [ ] enforce #[deny(arithmetic_overflow)]
• [ ] enforce clippy::integer_arithmetic
• [ ] disallow the use of f32 / f64 entirely