Define lightweight RHIME run/model specs and terminology

[brendan-m-murphy]

Goal

Introduce the minimum spec structures needed to configure standard RHIME and multi-sector RHIME
without adding keyword arguments to fixedbasisMCMC.

Tasks

• Add small dataclasses or typed structures for RhimeRunSpec, RhimeModelSpec, SectorSpec, and
  output/sampling settings.
• Use flux_sources in new config/API examples. Treat legacy emissions_name only as an old config
  compatibility key.
• Define source as the OpenGHG metadata key used to retrieve flux data.
• Define sector as the model component optimised separately.
• Define tracer as an additional species used to constrain the primary species, normally with
  linked forward models.
• Define species as the primary gas/tracer name used for object-store lookup and naming.

Acceptance criteria

• A one-sector spec represents the standard RHIME model.
• A multi-sector spec represents at least FF, GPP, TER, and Ocean with independent prior
  arguments.
• Specs can be constructed directly in Python without config-file parsing.
• Documentation does not introduce new uses of the word "emissions" except when describing legacy
  config keys.

[brendan-m-murphy]

PR #420 implements the first lightweight RHIME dataclasses (RhimeRunSpec, RhimeModelSpec, SectorSpec, RhimeOutputSpec, RhimeResult) and switches new public config/API names toward flux_sources and snake-case prior/output names.

Remaining work for this issue should focus on turning those lightweight structs into the backing model for config parsing and runner setup. _run_common is still doing too much manual normalization, validation, default handling, and mapping from config values to model/data/output requirements.

The current PR also revealed a concrete config-parser follow-up: malformed structured values such as sector_priors = {... can parse as strings and fail later with confusing runtime errors. That should be handled here or split into a small validation issue. (Note: issue #419 documents this issue.)