is54/uvselp: fix segfault by adding ANSI C prototypes

[ludomal]

Fix uvselp segfault on modern compilers (GCC 6+, Clang, MSVC)

Summary

Replace K&R-style empty-parentheses function declarations in src/is54/ with proper ANSI C prototypes. This fixes segfaults/crashes that occur with any modern compiler.

Note that bit-exact output verification tests are not included because floating-point rounding produces platform-dependent results differing by more than ±1. This is pre-existing (#24) and inherent to float arithmetic, and original tests vectors could not be verified on any tested platforms (AMR64 and x86).

This PR:

• Fixes uvselp: assert fails using MSVC #41
• Fixes UVSELP: segfaults using GCC 6+ or CLANG [TESTS] #132

Problem

The IS-54 VSELP codec uses 1990-era K&R function declarations:

/* In header files */
FTYPE EXCITE ();
FTYPE RS_RR ();
int G_QUANT ();

/* K&R function definition */
FTYPE RES_ENG (rq0, k)
     FTYPE rq0;
     FTYPE *k;
{ ... }

When FTYPE is float (the default), this triggers C default argument promotion: callers promote float arguments to double because no prototype is visible. But the function bodies expect float parameters. This ABI mismatch causes:

• GCC 6+ / Clang on x86-64: segfault in EXCITE. gsp0 receives a garbage value because register assignments are shifted by the size mismatch
• MSVC: assert(tmp >= 0) fails in RES_ENG (interp.c:134). rq0 receives corrupted data, producing negative values under sqrt
• ARM64 (Apple Silicon, aarch64): bus error (EXC_BAD_ACCESS). float is passed in S registers, double in D registers, completely different register banks

Fix

Added proper ANSI C prototypes to all header files and converted K&R function definitions to ANSI style:

File	Change
r_sub.h	Full prototypes for 7 functions
t_sub.h	Full prototypes for 12 functions
vselp.h	Full prototypes for 20 functions
vselp.c	Fix get_codes/put_codes function pointer declarations
interp.c	Convert RES_ENG K&R definition to ANSI; fix ATORC forward decl
makec.c	Convert A_SST K&R definition to ANSI; fix ATORC, ATOCOR, LEVINSON forward decls
g_quant.c	Fix corr forward declaration
lag.c	Fix i_resp forward declaration

The internal precision remains float (FTYPE). The fix only corrects the calling convention so arguments are passed in the correct registers/stack positions. The algorithm and arithmetic are unchanged.