[Lexical] Undocumented HexFloat syntax

[dlangBugzillaToGithub]

zxinsworld reported this on 2024-12-13T15:56:18Z

Transferred from https://issues.dlang.org/show_bug.cgi?id=24894

Description

This code compiles:
```d
void main(){
	float x = 0xp1;
}
```
However, the grammar in the spec doesn't list any way to write a HexFloat without leading digits:
```
HexFloat:
    HexPrefix HexDigitsNoSingleUS . HexDigitsNoStartingUS HexExponent
    HexPrefix . HexDigitsNoStartingUS HexExponent
    HexPrefix HexDigitsNoSingleUS HexExponent
```
So we need to either add ``HexPrefix HexExponent`` to the grammar, or deprecate``0xp1`` in case it wasn't meant to work in the first place.

[Prthmsh7]

#4211 This change documents the existing behavior of the D compiler, which already accepts the 0xp1 syntax. The value of a hex float literal in this form is interpreted as zero (0) shifted left by the exponent value. Rather than deprecating the syntax, I updated the specification.