Fix shift-exponent UB in bitstoint()

[iliasabk]

Summary

Fixes #415 — bitstoint() invokes undefined behavior for a component precision of 32:

• 1 << (prec - 1) = 1 << 31 overflows into the sign bit of a 32-bit int
• 1 << prec = 1 << 32 has an out-of-range shift exponent
• v - (1 << prec) then wraps in unsigned 32-bit, producing a wrong sign extension, and JAS_ONES(32) = (1 << 32) - 1 has the same UB — also reached via inttobits()

Fix

• Mask computation uses JAS_POW2_X(uint_fast64_t, prec) - 1 (defined for prec <= 32)
• The sign test uses JAS_POW2_X(uint_fast32_t, prec - 1) (bit 31 is fine in unsigned)
• The sign extension computes v - 2^prec in jas_seqent_t (64-bit), giving the mathematically correct negative result

Testing

Reproduced with the PoC attached to #415 under UBSan:

jas_image.c:1044:4: runtime error: left shift of 1 by 31 places cannot be represented in type 'int'
jas_image.c:1044:4: runtime error: signed integer overflow: -2147483648 - 1

After the patch: no sanitizer report, decode fails gracefully on the malformed input.