Add imputation MVP

[jpmcga]

• take a list of missing row/column ids (no automated detection)
• Only imputes for rows/columns with 'good' neighbors (ie. no consecutive missing)

Runs:

• D01355: missing row 37
• D01270: missing columns 10, 34
• D01254: 96x, missing row 30
• D01332: 96x, missing 31-35

[jpmcga]

We tested 'average of neighbors' algorithm with D01355, D01270; resulted in the imputed tixels clustering out with neighbors for tixels in heterogenous regions (cluster 1, below). Tixels in homogenous regions looked ok. Let's try two different approaches to resolve artifact clusters,

1. Remove diagonal neighbors from averaging so only direct neighbors used in imputations,
2. New algorithm: copy exactly the direct neighbor (no diagonal) with the higher fragment count.

@JoshuaBAtlas

[jpmcga]

Completed for single lane in https://github.com/atlasxomics/ATX_impute/tree/main