# fix(convert): zero-pad each channel in RgbToHex

## Problem

`RgbToHex` (and its alias `Rgb2hex`) does not zero-pad channel values below
`0x10`, so a single hex digit is emitted for those channels:

```go
color.RgbToHex([]int{1, 2, 3})   // "123"    , want "010203"
color.RgbToHex([]int{0, 0, 0})   // "000"    , want "000000"
color.RgbToHex([]int{10, 11, 12}) // "0a0b0c" -> actually "abc"
```

This breaks two contracts:

1. The round-trip with `HexToRgb` is no longer an identity for colors with any
   channel below 16:

   ```go
   color.RgbToHex(color.HexToRgb("010203")) // "123", not "010203"
   ```

2. `Rgb2basic` looks a color up in `hex2basicMap`, whose keys are all
   fixed-width 6-digit hex strings (e.g. "000000"). Because `RgbToHex(0,0,0)`
   returns "000", pure black never matches the fast-path map and silently falls
   through to the approximate `RgbToAnsi`.

The sibling converter `RgbTo256` already builds its hex key with
`fmt.Sprintf("%02x%02x%02x", ...)`, so this is an inconsistency within the same
file.

## Root cause

`convert.go`, `RgbToHex`:

```go
for _, v := range rgb {
	hexNodes = append(hexNodes, strconv.FormatInt(int64(v), 16))
}
```

`strconv.FormatInt(_, 16)` produces the minimal representation, i.e. one digit
for values 0-15.

## Fix

Format each channel as a fixed two-digit hex value:

```go
hexNodes = append(hexNodes, fmt.Sprintf("%02x", v))
```

`fmt` is already imported in the file. Existing values >= 16 are unchanged.

## Test

Extended `TestRgbToHex` in `utils_test.go` with small-channel cases
(`{0,0,0}`, `{1,2,3}`, `{10,11,12}`) and an explicit
`RgbToHex(HexToRgb(hex)) == hex` round-trip check.

- Gate: `go test ./` -> ok (PASS)
- `go vet ./...` -> exit 0
- `go build ./...` -> exit 0

Counterfactual: reverting `RgbToHex` to `strconv.FormatInt(int64(v), 16)` makes
the new test fail with `expect: "abc"` vs `actual: "0a0b0c"`.

Note: `gofmt -l convert.go` reports the file, but this is pre-existing (a doc
comment near line 866, unrelated to this change and present on the base commit).
The changed region is gofmt-clean, so unrelated lines were intentionally left
untouched.