Allow projections in find headers

[Shekar V (vshekar)]

This PR is specifically meant for the way data is stored in the MX beamlines, it can be expanded to other collections if needed.

The MX beamlines use analysis store headers to store the results of rasters. An example of a raster result is as follows:

{
    "time": 1722530001.5539453,
    "uid": "595e358c-b1a6-4d99-ac1b-70a92937cb77",
    "provenance": {
        "lsdc": 1
    },
    "result_type": "rasterResult",
    "owner": "lsdc-amx",
    "sample": "b8086557-e2cc-4466-994c-137ce90fafa4",
    "request": "924916a6-7530-424a-861a-9078c7e94985",
    "result_obj": {
        "sample_id": "b8086557-e2cc-4466-9e4c-127ce90fafa4",
        "parentReqID": "a39bfafd-2398-49db-9e40-977f07be4c32",
        "rasterCellMap": {
            "cellMap_1": {
                "x": 5662.5,
                "y": 700.9384045160338,
                "z": 233.98982510854614
            }, ... <hundreds more cellMap co-ordinates>
        }
        "rasterCellResults": {
                    "type": "dialsRasterResult",
                    "resultObj": [
                        {
                            "image": [
                                "/path/to/master.h5",
                                1
                            ],
                            "spot_count": 0.0,
                            "spot_count_no_ice": 0.0,
                            "d_min": 50.0,
                            "d_min_method_1": 50.0,
                            "d_min_method_2": 50.0,
                            "total_intensity": 0.0,
                            "cellMapKey": "cellMap_1"
                        }, ... <hundreds more cell results>
        "proposalID": "312346",
        "beamline": "amx"

For context, dumping a raster data like above with 195 cells into a file with indent=4 takes around 3500 lines. Not all of the data is necessary for purposes of reconstructing the raster. Specifically, there is no need for "rasterCellMap", or data within "rasterCellResults.resultObj" for certain situations. By using projections, one can remove keys like "rasterCellMap" and potentially reduce the processing time to get raster data, especially if we filter a large number of rasters in 1 query.
Removing keys from within "rasterCellResults" is not straightforward but could be, by changing the structure of data. But that is a PR for another repo.