`--processors` /` n_processors` greater than 1 _still_ produces incorrect token timestamps

[tefkah]

hi, i would like to flag #2036 as not being fixed, see here the following reproduction

when running

build/bin/whisper-cli -m models/ggml-large-v3-turbo.bin --output-file domecium --output-json-full --language en -f domecium.mp3 --processors 2

on this file from this short poetry collection

domecium.mp3

it mentions that it has been split around 1:35

whisper_full_parallel: the audio has been split into 2 chunks at the following times:
whisper_full_parallel: split 1 - 00:01:35.460
whisper_full_parallel: the transcription quality may be degraded near these boundaries

When we look at domecium.json, we see the token level timestamps reset around that boundary

			{
					"text": "[_TT_1266]",
					"timestamps": {
						"from": "00:01:35,460",
						"to": "00:01:35,460"
					},
					"offsets": {
						"from": 95460,
						"to": 95460
					},
					"id": 51631,
					"p": 0.95917,
					"t_dtw": -1
				}
			]
		},
		{
			"timestamps": {
				"from": "00:01:35,460",
				"to": "00:01:47,640"
			},
			"offsets": {
				"from": 95460,
				"to": 107640
			},
			"text": " in days bygone long gone my father's mother who is now blest with the blest would take me out to walk",
			"tokens": [
				{
					"text": "[_BEG_]",
					"timestamps": {
						"from": "00:00:00,000",
						"to": "00:00:00,000"
					},
					"offsets": {
						"from": 0,
						"to": 0
					},
					"id": 50365,
					"p": 0.997823,
					"t_dtw": -1
				},
				{
					"text": " in",
					"timestamps": {
						"from": "00:00:00,000",
						"to": "00:00:00,290"
					},
					"offsets": {
						"from": 0,
						"to": 290
					},
					"id": 294,
					"p": 0.75974,
					"t_dtw": -1
				},

So the token level timestamps still reset. The segment level timestamps are still correct however.