[FileFormats.MPS] improve performance of parsing each line

[odow]

Closes #2888

Before

julia> using JuMP

julia> filename = "/Users/odow/Downloads/supportcase19.mps"
"/Users/odow/Downloads/supportcase19.mps"

julia> @time read_from_file(filename)
  7.129035 seconds (43.28 M allocations: 4.544 GiB, 19.73% gc time)
supportcase19
├ solver: none
├ objective_sense: MIN_SENSE
│ └ objective_function_type: AffExpr
├ num_variables: 1429098
├ num_constraints: 4298007
│ ├ AffExpr in MOI.EqualTo{Float64}: 10351
│ ├ AffExpr in MOI.GreaterThan{Float64}: 362
│ ├ VariableRef in MOI.GreaterThan{Float64}: 1429098
│ ├ VariableRef in MOI.LessThan{Float64}: 1429098
│ └ VariableRef in MOI.Integer: 1429098
└ Names registered in the model: none

After

julia> using JuMP

julia> filename = "/Users/odow/Downloads/supportcase19.mps"
"/Users/odow/Downloads/supportcase19.mps"

julia> @time read_from_file(filename)
  3.720214 seconds (20.38 M allocations: 3.032 GiB, 30.93% gc time)
supportcase19
├ solver: none
├ objective_sense: MIN_SENSE
│ └ objective_function_type: AffExpr
├ num_variables: 1429098
├ num_constraints: 4298007
│ ├ AffExpr in MOI.EqualTo{Float64}: 10351
│ ├ AffExpr in MOI.GreaterThan{Float64}: 362
│ ├ VariableRef in MOI.GreaterThan{Float64}: 1429098
│ ├ VariableRef in MOI.LessThan{Float64}: 1429098
│ └ VariableRef in MOI.Integer: 1429098
└ Names registered in the model: none

[odow]

We're still slower than HiGHS (which itself isn't the fastest), but we also do a lot more stuff than HIGHS to build the Julia data structures to go from JuMP -> MOI -> HiGHS.

julia> filename = "/Users/odow/Downloads/supportcase19.mps"
"/Users/odow/Downloads/supportcase19.mps"

julia> using HiGHS

julia> highs = Highs_create()
Ptr{Nothing}(0x0000000122580c00)

julia> @time Highs_readModel(highs, filename)
Running HiGHS 1.13.0 (git hash: 1bce6d5c8): Copyright (c) 2026 under Apache 2.0 license terms
Using BLAS: blastrampoline 
  2.759003 seconds (1.51 k allocations: 74.609 KiB, 0.14% compilation time)