fix: add bounds check for star width/precision in std.format

[He-Pin]

Motivation

std.format("%*d", [5]) threw ArrayIndexOutOfBoundsException instead of a clean error. The bounds check at the top of the loop ran before star consumption but not after — once * width/precision consumed values and incremented i, the subsequent valuesArr.value(i) access had no guard.

Both go-jsonnet and jrsonnet report a clean "not enough values" error. Python raises TypeError: not enough arguments for format string.

Modification

Add i >= valuesArr.length guards after each i += 1 in the three star-consuming branches: (true,false), (false,true), and (true,true). The error message matches the existing pre-loop check format.

Result

Star format specs with insufficient values now produce a clean user-facing error instead of an internal crash.

Behavior comparison

Input	go-jsonnet	jrsonnet	Python	sjsonnet (before)	sjsonnet (after)
"%*d" % [5]	error: "Not enough values to format: 1, expected more than 1"	error: "not enough values"	TypeError: not enough arguments	AIOOBE crash	error: "too few values to format: 1, expected at least 2"
"%*.*d" % [5, 3]	error: "Not enough values to format: 2, expected more than 2"	error: "not enough values"	TypeError: not enough arguments	AIOOBE crash	error: "too few values to format: 2, expected at least 3"
"%*d" % [5, 42]	" 42"	" 42"	" 42"	" 42"	" 42" (unchanged)
"%*s" % []	error: "Not enough values"	error: "not enough values"	TypeError	error: "too few values to format: 0, expected at least 1"	(unchanged, caught by pre-loop check)

Test plan

• std.format("%*d", [5]) → clean error instead of AIOOBE
• std.format("%*.*d", [5, 3]) → clean error instead of AIOOBE
• std.format("%*d", [5, 42]) → " 42" (no regression)
• Full test suite passes (73/73)