read the metadata of the SAR image files

[keewis]

• towards different file formats #3

The SAR image files have a header, and then contain a (large) number of uncompressed data records. This only defines the structure of the SAR image header, because reading all data records at once would consume too much memory.

Todo: figure out the byte ranges for each of the data records.

[keewis]

with the latest commits, we can extract the metadata including byte offsets of the actual data, then separately read the data (this takes a 100-400ms or more, depending on network speed / disk access time).

Extracting the metadata also takes about a minute per sar image file (we have lots of those per dataset), so we might want to create cache files for those, reducing the time to a few seconds at most.