Support udecimal library in code generation

[michaelwilner]

The quagmt/udecimal library offers some striking performance gains over shopspring/decimal especially when parsing and serializing strings. Benchmarks from udecimal reflect 1-2 orders of magnitude improvement on these paths

goos: linux
goarch: amd64
pkg: github.com/quagmt/udecimal/benchmarks
cpu: Intel(R) Core(TM) i9-14900HX
                                                                                            │   shopspring   │                udecimal                │
                                                                                            │     sec/op     │    sec/op     vs base                  │
Parse/1234567890123456789.1234567890123456879-32                                               384.35n ± 10%   25.16n ±  5%  -93.45% (p=0.000 n=10)
Parse/123-32                                                                                   77.475n ± 13%   8.046n ±  8%  -89.61% (p=0.000 n=10)
Parse/123456.123456-32                                                                         128.35n ±  5%   13.25n ±  3%  -89.68% (p=0.000 n=10)
Parse/1234567890-32                                                                             82.47n ±  7%   11.20n ±  1%  -86.42% (p=0.000 n=10)
Parse/0.1234567890123456879-32                                                                 262.30n ± 17%   17.89n ±  4%  -93.18% (p=0.000 n=10)
Parse/12345678901234567890123456789.123-32                                                     312.20n ±  8%   48.83n ±  3%  -84.36% (p=0.000 n=10)
ParseFallBack/123456789123456789123456.1234567890123456-32                                      373.2n ± 13%   337.6n ±  2%        ~ (p=0.105 n=10)
ParseFallBack/111222333444555666777888999.1234567890123456789-32                                418.5n ±  5%   325.9n ±  5%  -22.12% (p=0.000 n=10)
String/1234567890123456789.1234567890123456879-32                                              284.45n ± 26%   53.19n ±  3%  -81.30% (p=0.000 n=10)
String/123-32                                                                                  110.55n ±  7%   14.72n ±  2%  -86.68% (p=0.000 n=10)
String/123456.123456-32                                                                        133.00n ± 10%   27.95n ±  6%  -78.99% (p=0.000 n=10)
String/1234567890-32                                                                           145.45n ± 18%   24.37n ±  3%  -83.25% (p=0.000 n=10)
String/0.1234567890123456879-32                                                                173.80n ± 15%   34.36n ± 11%  -80.23% (p=0.000 n=10)
String/12345.1234567890123456789-32                                                            241.90n ±  4%   37.68n ±  5%  -84.42% (p=0.000 n=10)
StringFallBack/123456789123456789123456.1234567890123456-32                                     394.1n ± 15%   174.1n ±  3%  -55.82% (p=0.000 n=10)
StringFallBack/111222333444555666777888999.1234567890123456789-32                               353.4n ± 26%   174.4n ±  8%  -50.65% (p=0.000 n=10)

Performance compared between the FIXUDecimal and FIXDecimal types:

goos: darwin
goarch: arm64
pkg: github.com/quickfixgo/quickfix
cpu: Apple M3 Pro
BenchmarkFIXUDecimalRead
BenchmarkFIXUDecimalRead-11     	53151679	        22.33 ns/op
BenchmarkFIXUDecimalWrite
BenchmarkFIXUDecimalWrite-11    	29959304	        39.36 ns/op
BenchmarkFIXDecimalRead
BenchmarkFIXDecimalRead-11     	13141797	        79.57 ns/op
BenchmarkFIXDecimalWrite
BenchmarkFIXDecimalWrite-11    	 5214294	       235.6 ns/op