feat: compute parametes are now optional

  `service_order.source_protocol`. It fails this rule, but splitting it means defining the value

  vocabulary for both columns, which is a design decision rather than a naming fix. Tracked

  separately.

- OEG's `SRMComputeIntent` diverges structurally from SRM's canonical `ComputeRequirements`

  (`srm.compute/v1`): `standalone` sits inside `compute` rather than at the top level;

  `cpu_millicores`, `memory_mb`, `topology` and `StorageVolume.mount_point` are optional in OEG

  and required in SRM; and OEG types as `str` several fields SRM types as enums

  (`AcceleratorType`, `InterfaceVisibility`, `TransportProtocol`, `RuntimeKind`). These are shape

  mismatches rather than naming, and are fixed separately.

- OEG's `SRMComputeIntent` still diverges from SRM's canonical `ComputeRequirements`

  (`srm.compute/v1`) on remaining shape mismatches: `StorageVolume.mount_point` is optional in

  OEG and required in SRM, and OEG types as `str` several fields SRM types as enums

  (`AcceleratorType`, `InterfaceVisibility`, `TransportProtocol`, `RuntimeKind`).

  `cpu_millicores`, `memory_mb`, and `topology` are optional on both sides.

- `SRMCapabilityRequirement.deployment_unit_ref` is retained. SRM's column is

  `deployment_unit_id`, but the catalog POST creates units and requirements in one transaction,

  so the requirement can only reference a unit that has no id yet. `ref` is an SRM column name

  "schema_version": "srm.compute/v1",

  "compute": {

    "cpu_millicores": 2000,              // CAMARA cpuPool.numCPU x1000 / GSMA numCPU

    "memory_mb": 2048,                   // CAMARA cpuPool.memory / GSMA memory

    "cpu_millicores": 2000,              // CAMARA cpuPool.numCPU x1000 / GSMA numCPU;

                                         // omit when the northbound manifest has no CPU sizing

    "memory_mb": 2048,                   // CAMARA cpuPool.memory / GSMA memory;

                                         // omit when the northbound manifest has no memory sizing

    "accelerator": {                     // present only if GPU/VPU/FPGA required

      "type": "gpu",                     // gpu | vpu | fpga

      "units": 1,                        // CAMARA gpuPool.numGPU / GSMA numGPU

    ]

  },

  // placement/topology constraints (CAMARA topology.* / gpuPool.topology.*)

  // placement/topology constraints (CAMARA topology.* / gpuPool.topology.*);

  // omit the whole object when the northbound manifest has no topology

  "topology": {

    "min_nodes": 2,

    "min_node_cpu_millicores": 1000,