dvc exp run: Unnecessary "Collecting and computing hashes" after chaging cmd in dvc.yaml

[aljeshishe]

Bug Report

Description

Unnecessary "Collecting and computing hashes" after changing cmd in dvc.yaml

Reproduce

mkdir repro 
cd repro
git init 
dvc init
mkdir 1
mkdir 2
cd 1
dd if=/dev/random of=./data  bs=10GB count=1
dvc add data
cd ../2
dd if=/dev/random of=./data  bs=10GB count=1
dvc add data 
cd ..
git add .dvc
git commit -m init

cat <<EOF > dvc.yaml
stages:
    target:
        cmd:
            - sleep 1
        deps:
            - 1/data

EOF
dvc exp run 

cat <<EOF > dvc.yaml
stages:
    target:
        cmd:
            - sleep 1
        deps:
            - 2/data

EOF
dvc exp run 

Expected

I expect that hash is already calculated for all files and there is no Collecting and computing hashes stage

Environment information

Output of dvc doctor:

dvc doctor
DVC version: 3.44.0 (pip)
-------------------------
Platform: Python 3.10.12 on Linux-5.15.0-1053-aws-x86_64-with-glibc2.31
Subprojects:
	dvc_data = 3.11.0
	dvc_objects = 5.0.0
	dvc_render = 1.0.1
	dvc_task = 0.3.0
	scmrepo = 3.1.0
Supports:
	http (aiohttp = 3.9.3, aiohttp-retry = 2.8.3),
	https (aiohttp = 3.9.3, aiohttp-retry = 2.8.3)
Config:
	Global: /home/ubuntu/.config/dvc
	System: /etc/xdg/dvc

Additional Information (if any):
Some debugging showed following:

1. 2/data could not be located in dvc.lock/deps in StageLoader.fill_from_lock on 2nd dvc exp run
2. _compute_hash_info_from_meta(None) is called
3. _compute_hash_info_from_meta returns md5-dos2unix
4. That is why dvc doesn't use precomputed hash in following code:

def hash_file(
    path: "AnyFSPath",
    fs: "FileSystem",
    name: str,
    state: Optional["StateBase"] = None,
    callback: Optional["Callback"] = None,
    info: Optional[dict] = None,
) -> Tuple["Meta", "HashInfo"]:
    if state:
        meta, hash_info = state.get(path, fs, info=info)
        if hash_info and hash_info.name == name:
            return meta, hash_info