[BUG] False negative result of generated test with union for one of init value

[tyuldashev]

Description
With looking identical initialization UTBot test gives different result of the test

To Reproduce

1. Run test generation on following example

// extracted from integration-tests/c-example/lib/structures/simple_unions.c
union MainUnion {
    union InnerUnion {
        char c;
        short s;
    } inner;

    int x;
    long long y;
};

signed char operate_with_inner_unions(union MainUnion st) {
    if (st.x == 5 || st.y == 5 || st.inner.c == '5' ||
        st.inner.s == 5 ) {
        return '5';
    }

    if (st.x == 5 || st.y == 102 || st.inner.s == 15) {
        return st.inner.c;
    }

    return 'o';
}

2. One of resulting tests looks like:

TEST(regression, operate_with_inner_unions_test_3)
{
    signed char actual = operate_with_inner_unions({
        // .inner = {
            // .c = '\x0f'
            // .s = 15
        // }
        // .x = 15
        .y = 15LL
    });
    EXPECT_EQ('\x0f', actual);
}

3. Comment initialization through .y and uncomment initialization with .s, so test becomes

TEST(regression, operate_with_inner_unions_test_3)
{
    signed char actual = operate_with_inner_unions({
        .inner = {
            // .c = '\x0f'
            .s = 15
        }
        // .x = 15
        // .y = 15LL
    });
    EXPECT_EQ('\x0f', actual);
}

4. Run the test - it still green as expected.
5. Comment line with .s and uncomment with .c, so test becomes

TEST(regression, operate_with_inner_unions_test_3)
{
    signed char actual = operate_with_inner_unions({
        .inner = {
            .c = '\x0f'
            // .s = 15
        }
        // .x = 15
        // .y = 15LL
    });
    EXPECT_EQ('\x0f', actual);
}

7. Run the test again, now it fails with error:

Expected equality of these values:
  '\x0f'
    Which is: '\xF' (15)
  actual
    Which is: 'o' (111, 0x6F)

Expected that test still passes

[alexey-utkin]

@tyuldashev that is the predictable case that shows the record into alignment gape.
That is the reason why sometime we need to initialize init the union from the stream bytes , if we don't find "field path" to init the structure in the desired state. Currently the longest "field path" is used for initialization.
Umbrella fix from
[BUG] Anonymous union initialized using 'from_bytes' in some cases #447 (#467)