tccelf: fix int(0x8000_0000) overflow in shf_private constant

[quaesitor-scientiam]

Problem

src/tccelf.v has two constants using explicit int() casts on hex literals:

const shf_private = int(0x8000_0000)  // overflows int (2147483648 > 2147483647)
const shf_dynsym  = int(0x4000_0000)  // doesn't overflow, but cast is redundant

shf_private overflows V's int. V currently emits a warning and will make this a hard error soon. The overflow causes broken generated C that TCC rejects:

src/tccelf.v:15:21: warning: value `0x80000000` overflows `int`, this will be considered hard error soon
cc: error: invalid operand types for binary operation

This makes vtcc fail to build in the vlang/v CI with V panic: result not set (failed). Tracked at vlang/v#26853.

Fix

Two commits:

1. shf_private: replace int(0x8000_0000) with -2147483648 (INT_MIN — same bit pattern in two's complement, type stays int, no cascading changes needed in new_section/new_symtab/struct Section.sh_flags)

2. shf_dynsym: remove the redundant int() cast — the value 0x4000_0000 = 1073741824 fits int fine, the cast is just noise

-const shf_private = int(0x8000_0000)
-const shf_dynsym  = int(0x4000_0000)
+const shf_private = -2147483648 // = 0x8000_0000 as two's complement int (INT_MIN)
+const shf_dynsym  = 0x4000_0000 // remove redundant int() cast; value fits int

Fixes #6