RF/FIX: Prioritize sbref and shortest echo for SyN-SDC

[effigies]

This PR resolves the twin issues of either creating an estimator per echo or including all echos (with different contrast and dropout properties) in multi-echo SyN.

Because we want the shortest echo, the simplest way to do this was to sort all possible EPIs by suffix and echo time, to push the shortest echo sbref to the front of the line, or the shortest echo bold/dwi, failing that. We then set "IntendedFor" to the set of sbrefs/echos matching all other entities/path components. Keeping track of what files have been included in an "IntendedFor" avoids finding multiple estimators for a set.

New estimator results:

❯ sdcflows-find-estimators --subject 02 --fmapless . 
Estimation for </data/openneuro/ds000210> complete. Found:
	sub-02
		FieldmapEstimation(sources=<2 files>, method=<EstimatorType.ANAT: 5>, bids_id='auto_00000')
			None	anat/sub-02_T1w.nii.gz
			j	func/sub-02_task-cuedSGT_run-01_echo-1_bold.nii.gz
		FieldmapEstimation(sources=<2 files>, method=<EstimatorType.ANAT: 5>, bids_id='auto_00001')
			None	anat/sub-02_T1w.nii.gz
			j	func/sub-02_task-cuedSGT_run-02_echo-1_bold.nii.gz
		FieldmapEstimation(sources=<2 files>, method=<EstimatorType.ANAT: 5>, bids_id='auto_00002')
			None	anat/sub-02_T1w.nii.gz
			j	func/sub-02_task-cuedSGT_run-03_echo-1_bold.nii.gz
		FieldmapEstimation(sources=<2 files>, method=<EstimatorType.ANAT: 5>, bids_id='auto_00003')
			None	anat/sub-02_T1w.nii.gz
			j	func/sub-02_task-cuedSGT_run-04_echo-1_bold.nii.gz

Closes #363.

[effigies]

I should probably add similar tests for multiecho pepolar. Can do that in this or a separate PR.

@mgxd Sorry to be bugging you, but I'd really like to get fMRIPrep out this week if I can, and this is a blocker. If you have time, I'd appreciate a speedy review.

[mgxd]

LGTM- this functionality seems self-contained enough to potentially be a separate function

[effigies]

Not a massive refactor, but at least it encapsulated it. WDYT?

[mgxd]

Looks clean