Add a "use float" pragma [feature]

[richardleach]

Perl has an integer pragma, which directs the compiler to use integer operations from the point of use to the end of the enclosing BLOCK. Besides the user-visible behaviour, described at https://perldoc.perl.org/integer.html, perl will then use integer-specific versions of various operators. (e.g. pp_i_add, rather than the general purpose pp_add.)

A useful feature might be a corresponding float pragma, to force the compiler to use floating point operations.

This would entail the likes of:

• Modifying the parser/lexer (?) to recognise the pragma
• Adding new floating point ops
• Adding new tests
• All the fixups and regens that go with the above.

Benefits to users who want this:

• Improved operator performance (no attempts to first convert operands to IV/UV values before falling back to floating point operations)
• Fewer unnecessary upgrades of the underlying SV (see IV / NV upgrades often seem unnecessary and costly #17777 for some discussion)